← Home Photos Archive About Reading Also on Micro.blog
  • Found some time today to upload the Storyform of The Abyss into Subtxt. Theory and Hunch are at the center of Bud and Lindsey’s argument. The world on the brink of Armageddon mixed with a Be-er Main Character/leader is more than just a little similar to Crimson Tide. One of my all-time favorites.

    → 8:27 PM, Dec 31
  • The Abyss in 4K at home…😍

    → 2:42 AM, Dec 31
  • I see articles like this posted on Bluesky, and I so badly would love to share all the positive experiences I’ve had writing with AI and of users enjoying their experiences with the same… but then I read the comments and everyone is so virulently anti-AI that I just retreat back to the comfort of my own experience and post my little bit into the void.

    Even as writers battle the scourge of A.I., many have begun to use it as a tool for making sentences. More than that, some have embraced A.I. as the latest iteration of an ancient literary conceit: the fantasy of a co-author, a confidant, a muse — an extra intelligence, a supplemental mental database. Poets and novelists once turned to séances, Ouija boards and automatic writing for inspiration. Now they can summon a chatbot to their laptops.

    → 2:47 PM, Dec 30
  • One of my favorite discoveries of 2023 was Thomas Flight. His channel gives me hope that cinema is not dead. His Best Films of 2023 are a good reference of what was good.

    And if you’re like me, and would rather just see the film list without any clips or criticism first, here they are:

    • Dream Scenario
    • Anatomy of a Fall
    • Fallen Leaves
    • Past Lives
    • May December
    • Godland
    • The Holdovers
    • Poor Things
    • The Boy and The Heron
    • Asteroid City

    Out of all of these, I think Flower Moon is the only one with a complete story (haven’t seen them all yet, though). Past Lives was beautiful, but wasn’t all there in terms of narrative.

    Looking forward to seeing the others. 😃

    → 12:49 PM, Dec 30
  • A balanced non-panicky take on copyright and LLMs. Love how it calls The NY Times out for what I’ve always known: the only artists afraid of AI are deeply insecure of their own self-worth:

    Let me let you in on a little secret: if you think that generative AI can do serious journalism better than a massive organization with a huge number of reporters, then, um, you deserve to go out of business. For all the puffery about the amazing work of the NY Times, this seems to suggest that it can easily be replaced by an auto-complete machine.

    → 12:17 PM, Dec 28
  • The NYTimes suing OpenAI and Apple courting NYTimes for AI are totally not connected at all.

    → 1:00 PM, Dec 27
  • Finally saw Killers of the Flower Moon this week. While extremely long, I was never bored and quite caught up in the drama of it all. After a couple of days, I believe I understand the Storyform it all. Building it now…

    → 8:28 PM, Dec 23
  • Took me all day, but FINALLY got parallel tool calling with OpenAI to work! 🎉

    → 8:38 PM, Dec 22
  • Sorry Substack people, I’m out.

    → 4:25 PM, Dec 22
  • Daily appreciation of the greatest comic book artist ever, D.M.

    → 1:24 PM, Dec 22
  • Bluesky - the “billionaire-proof” social network. Yay! Looking forward to permanently dropping all billionaire-networks in 2024. :)

    → 12:56 PM, Dec 22
  • Just discovered the Black Rain OST Hans Zimmer greatness from 1989. At 30-minutes we get the original motif that eventually became “Molossus” in Batman Begins. It’s like discovering a brand new cut of Zimmer’s Batman themes!

    → 1:57 PM, Dec 19
  • Our long national nightmare of widowed headlines is officially over. Tailwind adds text-wrap

    → 1:08 PM, Dec 19
  • Subtxt - the only predictive narrative framework that compares Star Wars to Good Will Hunting to the obscure, yet powerful, Quo Vadis, Aida? Nothing else like it.

    → 7:37 PM, Dec 18
  • Just signed up for DeepLearning.ai’s Retrieval Augmentation Course. Interested to see how much of it I’m already doing and what I can do to improve things even more.

    → 5:18 PM, Dec 17
  • 2024:

    It is difficult to get a man to understand something when his salary depends on his not understanding i

    → 12:02 PM, Dec 17
  • Funny. I read this quote from the weak-to-strong paper:

    We cannot directly study this problem today, but we can study a simple analogy: can small models supervise larger models (right)?

    As: “We can use small models to supervise larger models. Right?“

    → 11:03 AM, Dec 17
  • ChatGPT is suddenly way better again (can confirm) and everyone thinks they “fixed” something. It’s so obvious they downgraded it so they could train the next model based on the October breakthrough. They didn’t “find” gpus, they just stopped burning through the ones they already had.

    → 7:52 PM, Dec 16
  • A nice gentleman recently posted about Subtxt’s pricing structure. It was a wonderful opportunity to discuss the value add that comes from building your own AI chatbot (ie, you don’t have to worry about ChatGPT consuming everything).

    → 10:50 AM, Dec 16
  • Working a bit with multiple tool calls within a single request this evening (OpenAI). Floored by the results.

    → 1:32 AM, Dec 16
  • The key to a successful refactor is having the right team.

    → 3:14 PM, Dec 13
  • Refactoring refactoring factoringre toringrefac ingrefactor refactoring.

    → 2:40 PM, Dec 13
  • Looks like they finished training Q*/GPT-5 as Plus subscriptions are back online. 😃

    → 11:50 AM, Dec 13
  • Building Dream Machines - the latest on AI tech and advanced storytelling from Narrative First. ✨

    → 7:51 PM, Dec 11
  • Subtxt must be the only GPT-4 that doesn’t suffer from SAD (seasonal affective disorder). Asked it to help outline a scene, and it went above and beyond and wrote out the whole thing for me.

    → 5:56 PM, Dec 11
  • Updating all my function calls to “tool choice” calls because it’s OpenAI and it always changes. Felt like I was behind six months in doing this—but it’s only been 5 weeks! 💫😵‍💫

    → 7:00 PM, Dec 10
  • Learning about recursion tonight and it turns out that’s how Subtxt works! A story function that calls itself as it writes each Act, Sequence, and Scene. Apparently you need depth and hierarchy as well, and we’ve got that too!

    → 1:05 AM, Dec 10
  • As I start to build out all the content for Six-Figure AI App, I’m realizing that I’ve learned quite a bit over the past couple of years. Going to have to put in some extra effort and time to get it all down. 😅

    → 4:21 PM, Dec 9
  • Overheard about 2024:

    A society of people that are willing to understand on their own to create things, is a lot more powerful, and that’s the jackpot. A society that wants to pass the buck off, and not do the work, and be frustrated constantly, and not learn these things, whatever it might be…isn’t going to make it.

    Couldn’t agree more.

    → 1:58 PM, Dec 9
  • Andrej Karpathy on “hallucinations” being a feature, not a bug. I’ve always been surprised how many writers get annoyed when Subtxt sometimes get things “wrong”. That’s creativity!

    At the other end of the extreme consider a search engine. It takes the prompt and just returns one of the most similar “training documents” it has in its database, verbatim. You could say that this search engine has a “creativity problem” - it will never respond with something new. An LLM is 100% dreaming and has the hallucination problem. A search engine is 0% dreaming and has the creativity problem.

    When it comes to writing stories, I’ll take the hallucination problem over the creativity problem any day of the week.

    → 12:11 AM, Dec 9
  • Beyond excited about this new page in the guide - didn’t realize there were four (!) different AI tools in Subtxt: Brainstorming AI, Develop AI, Teach AI, and Subtxt AI. Subtxt AI Tools

    → 11:35 PM, Dec 8
  • Cleaning up documentation for Subtxt and there are some features that I installed a year ago that I totally forgot about. This thing is awesome! 💫

    → 11:11 PM, Dec 8
  • Beautiful new notebooks for writers - from iA. Not available yet, but you can sign up and with enough interest we can all get them sooner than later.

    → 1:59 PM, Dec 8
  • Having way too much fun putting this all together. 😃

    → 7:59 PM, Dec 7
  • New version of the Subtxt Guide is now up. Way simpler, way clearer, with tons of examples (most people had no idea what they could even do with Subtxt). Looking forward to filling it out throughout the rest of December.

    → 6:04 PM, Dec 7
  • Every time I open this, I feel like OpenAI is flipping me off. 😄

    → 4:08 PM, Dec 7
  • Best part about having clients/users all over the world? They send you local coffee beans from Colombia that are 5,000,000x better than anything you could get locally here in California. ☕️

    → 11:47 AM, Dec 7
  • Fascinating read on non-deterministic responses from GPT-4, even with temp set to 0. Was going to explore the new seed parameter soon, but perhaps it’s not ready for prime time yet either?

    → 1:59 AM, Dec 7
  • Finally got myself all set up with those fancy shop links at the top of my Twitter profile.

    → 2:34 PM, Dec 6
  • Found this YouTube video: AI Apps Making $20,000+ per month with 1 person teams and spent the whole morning answering all kinds of comments, basically confirming once again the need for Six-Figure AI App. Going to use all my responses for a FAQ on the eventual landing page. 😃

    → 12:54 PM, Dec 6
  • Seventeen years ago I started blogging about story structure and a particular theory of story I found really fascinating: Dramatica. I still remember my first post in 2006. It was an analysis of Spielberg’s “War of the Worlds” that I felt compelled to post after being invited to see it in his personal screening room at Amblin.

    I thought the movie was awful, and hoped that if I could just explain the theory to one person I could somehow change the future of storytelling.

    Fast forward to 2023 and now I work daily on an AI-integrated app that helps thousands of storytellers across the world with better stories.

    And now I want to stay writing about building profitable AI applications.

    Here we go.

    → 12:17 AM, Dec 6
  • Starting to see more and more of this (Building SaaS in a Weekend), which only confirms my motivation to keep working on Six-Figure AI App. Either I’m seeing more of it because I’m thinking of it, or this is the trend for 2024 - either way, going to be a fun year. 😊

    → 12:01 PM, Dec 5
  • Funny to see ‘text-davinci-003’ getting sunsetted by Jan. 4th. I remember when that was an upgrade!

    → 2:49 PM, Dec 4
  • Building something wish I had six years ago when I started building Subtxt: Six-Figure AI App. Everything you could possibly know about creating your own application, even if you know zero about modern coding. Designed to save you from the frustration of looking everywhere for answers.

    → 3:55 PM, Dec 3
  • Pretty sure I really enjoyed Oppenheimer.

    → 2:32 AM, Dec 3
  • Laravel Pulse is aptly named - Subtxt feels alive!

    → 12:41 PM, Dec 1
  • Wrote a little bit about iA’s new release of iA Writer 7. Basically, just wanted to test out what it’s like seeing your work elevated above the AI text within the doc. Gotta say, it’s a great feeling and welcomed experience to write again.

    → 11:45 AM, Dec 1
  • Absolutely love IA writer’s approach to integrating AI…all I can think about it how to clone this into Subtxt. 😅 But seriously, love that my favorite app to do writing in, has taken the whole AI thing seriously and keeps the integrity of the author intact. ❤️

    → 4:46 PM, Nov 30
  • Only bummer with this announcement from OpenAI is the lack of clarity on Ilya’s role going forward. Always felt like he was the heart of it all, but then again, i’m just on the outside.

    → 7:55 PM, Nov 29
  • Forgot to upgrade to the Whisper v3! Be interesting to see if there is any noticeable difference. Looking forward to when it will be part of the API.

    → 6:12 PM, Nov 29
  • Forgot to upgrade to the Whisper v3! Be interesting to see if there is any noticeable difference. Looking forward to when it will be part of the API.

    → 6:12 PM, Nov 29
  • It’s not going to be enough to simply upload transcripts and hope for the best. You’re gonna need someone in there to curate them and transcribe them, and make sure they stay aligned with all the information gathered over the decades. 😅

    → 8:02 PM, Nov 28
  • Discovered AJR last week, and found two songs that were GREAT introductions to really deep narrative concepts: inequity and justification. These two basically power all of Subtxt 😃 AJR, Inequity, and Justification

    → 3:51 PM, Nov 28
  • Maven Monday is today, which means you can 25% off a bunch of courses there (mine is one of the chosen ones 😊). Find out what 120 other students have so far (i.e., that story is so much more than you think it is!) Write Great Stories with AI 👉

    → 1:57 PM, Nov 27
  • After watching Karpathy’s new intro video to LLMs it’s apparent why prompt engineering isn’t a thing: you’re just training intuition in a System 1 mind (what we’d call Preconscious in story). The next step, likely all this Q* stuff, will be an API for a System 2 mind (Conscious) where–regardless of prompt–the LLM will do off and think through the question/problem step by step and then work out the best answer using test time compute. 🤓

    → 9:57 AM, Nov 26
  • 24 hours from concept to live site, complete with presale payment-link! Introducing Six-Figure AI App.

    → 8:59 PM, Nov 25
  • Had some fun last night talking to Muse about the implications of re-arranging the order of the Ghosts in A Christmas Carol. Given Subtxt’s narrative framework, Dickens got it wrong…😬 Fixing “A Christmas Carol” with Subtxt AI

    → 2:15 PM, Nov 25
  • On copyright and plagiarism:

    “This is nonsensical,” he wrote in the order. “There is no way to understand the LLaMA models themselves as a recasting or adaptation of any of the plaintiffs’ books.”

    Obviously the judge took 30 minutes out of his day to watch Karpathy. 😄

    → 12:31 PM, Nov 25
  • Watching Karpathy’s intro to LLMs. Great for those all new to this stuff, just learning how to use it for their work.

    → 11:38 AM, Nov 25
  • Finally. This long nightmare of stress is over. Everyone back at OpenAI, with just one major change: everyone is 2000x more fired up than a week ago.🚀

    → 12:47 AM, Nov 22
  • Hah! After outlining a story with some new algorithms, I asked a Persona based on me to actually write the story (using the extended context available to GPT-4 Turbo). It didn’t get very far. As you can see, it was a little too aligned with my laziness! 😄

    → 8:45 PM, Nov 21
  • Cannot fully describe in words how much I absolutely love this song.

    → 5:33 PM, Nov 21
  • First time doing the “I do, We do, They do” approach for teaching - SO MUCH more fun and the two hours went by in a flash! Convinced this is the way forward for 2024!

    → 1:57 PM, Nov 21
  • Fascinating image here on why using up all the context when working with AI is not always a good thing. That’s also why the Storyform approach used in Subtxt is so successful for dealing with longer narratives. We can use the Storyform to maintain thematic integrity by “chunking” the story down into manageable 16k chunks.

    → 11:22 AM, Nov 21
  • Watching DevDay again just to send out good vibes.

    → 8:43 PM, Nov 20
  • The fact that the employees behind the company that serves Subtxt’s AI services (OpenAI) are doing double-shifts during a company holiday to keep everything stable just goes to show I made the right choice going with them.

    → 12:07 PM, Nov 20
  • This 48 hour rollercoaster just came back into the station.

    → 1:21 AM, Nov 20
  • Thank God. Does this mean I can go to sleep now?

    → 1:09 AM, Nov 20
  • Unbelievable.

    → 10:39 PM, Nov 19
  • Iconic.

    → 2:24 PM, Nov 19
  • Woah. Sam is back tomorrow AND a 5-day fighting freeze. ✌️

    → 11:46 PM, Nov 18
  • This is what happens when you get so motivated by external events, that you rollout exponentially better writing tools for storytellers!

    → 8:13 PM, Nov 18
  • Looks like we already turned a corner on this - board was supposed to resign by 5pm, but missed the deadline. Sure it’ll all be fixed come Monday morning 9am. Altman/Brockman to return

    → 6:54 PM, Nov 18
  • Going to go to sleep and wake up tomorrow and find out that it was all a bad dream.

    → 2:10 AM, Nov 18
  • Proof once again: never ever work for anyone else. Even IF they’re not paying you a thing.

    → 4:42 PM, Nov 17
  • Best part about ChatGPT getting long-term memory like this? It runs on the Assistant API, which means we’ll have eventual access to our as well.

    • never have to explain yourself twice—ChatGPT will remember who you are, and all of the important people in your life
    • use ChatGPT as a repository for notes, ideas, quotes and more.
    • it will resurface all of these in new chats where relevant. A true second brain.

    → 10:05 AM, Nov 16
  • WONDERFUL tutorial on using the new Assistants API from OpenAI in Laravel. Still going to hold off until streaming is ready, but can’t wait to start integrating all of this goodness.

    → 12:43 AM, Nov 16
  • OK. Woah. I think I may have just unlocked some craziness using OpenAI’s latest models. Like insanely magical and exciting. Lots of work ahead!

    → 11:08 PM, Nov 15
  • Love this new setup. You can just reference a character you like in a story, and if Subtxt has access to the deep narrative structure, will give you thoughtful insights into what you might want to consider if you want to write something similar, or something inspired by that character.

    → 6:44 PM, Nov 15
  • So, apparently I can just use Laravel Scout to search my models more “fuzzy-like” (for mis-spellings) instead of trying to hobble together some LIKE within the Eloquent ‘where” models…six years, and I’m just realizing this!

    → 6:40 PM, Nov 15
  • Made a fun little GPT called “Subtx Muse, Jr.” who is a very much paired-down version of Subtxt Muse, but one that I hope gives users a chance to see what it’s like developing a story with it. Give it a whirl!

    → 2:32 PM, Nov 14
  • When someone sends me proof that other GPTs can do the same thing that Subtxt can, but it requires copying and pasting in knowledge from Subtxt. 🙄

    → 11:24 AM, Nov 14
  • Talk about looking “with better eyes”, this looks phenomenal! The Abyss in 4K

    → 10:32 PM, Nov 13
  • Upgraded everything in Subtxt to OpenAI’s new JSON response format. SO NICE finally being able to pull out all the hacky regex and string interpretations I’ve had running for the past year. Looking forward to zero bugs! 🪲❌

    → 9:48 PM, Nov 13
  • Ack! They’re starting to rollout videos from last week’s DevDay conference. Can’t wait to dive into these. First one is maximizing LLM Performance.

    → 6:20 PM, Nov 13
  • I know this is supposed to be some sort of anti-AI thing, but all I see is, “Look, AI can make something out of your pile of bad ideas!” 😄

    → 2:15 PM, Nov 13
  • What happens when you get 10/10 feedback from students in your last cohort? Why, your current rating notches up to 4.9! New one starts in a couple weeks: Write Great Stories with AI

    → 12:21 PM, Nov 11
  • Feeling like I should gift Dave Filoni a subscription to Subtxt … (full conversation here)

    → 7:05 PM, Nov 10
  • OpenAI’s new model updates have really taken Subtxt Muse to the next level. Company blog post on Exploring AI’s Role in Evolving Storytelling with Subtxt

    → 3:38 PM, Nov 10
  • 😂

    → 6:08 PM, Nov 9
  • Let’s go!

    → 3:07 PM, Nov 9
  • Paul Graham on AI businesses:

    Two ways for an AI company to protect itself from competition: (a) depend not just on AI but also deep domain knowledge about a particular field, (b) have a very close relationship with the end users.

    Done and done. 😊

    → 8:50 AM, Nov 9
  • So frustrating. Wife got the new GPTs before me. She doesn’t even use these things!!!!!

    → 7:32 PM, Nov 8
  • What the what?! Pinecone comes out with Canopy, a RAG framework, right when I was just getting comfortable with the new Assistants API process LOL.

    → 6:25 PM, Nov 8
  • Coming down from Monday’s event, but still reeling from some of the responses coming back from their API today. Have to re-think the possibilities of what is possible–which is quite a bit more than it was before Monday.

    → 5:03 PM, Nov 8
  • Glad to be a part of the “exponential” 20s.

    → 1:17 AM, Nov 8
  • Right. So I don’t know what’s real anymore. I’m trying to test out more reliable responses from OpenAI, and not only does it do that, but it’s proactively going above and beyond the questions I’m asking to give insane value to the end user. 💫 😵‍💫

    → 7:50 PM, Nov 7
  • And just like that 10 months of hacky-JSON errors ends with the very convenient response_format key. 🙏

    → 7:25 PM, Nov 7
  • Blown away yet again in the midst of teaching a cohort this afternoon. Asked one question about the Relationship Story Throughline and the response back was a picture perfect, no he said/she said illustration of the growth of the key relationship in a complete story. Only change? Model version #.

    It even went so far as to suggest Next Steps.

    What?!

    → 4:46 PM, Nov 7
  • Notes on OpenAI’s Retrieval Process

    Boris (tech engineer at OpenAI) on that retrieval slide:

    There’s so much to be said about this, but key is a good evaluation framework, and then doubling down on understanding what went wrong. Possibilities: 1. Outdated, inaccurate or contradictory information in source documents 2. Retrieval failure 3. Answer synthesis 4. Unclear query 5. Bad evaluation by a human reviewer

    It’s a definite art, but fun to engineer.

    More on retrieval (Yi Ding):

    A few thoughts on how OpenAI is implementing RAG in the new Assistants Retrieval tool before I was locked out.

    1. They’re splitting on newlines. You can tell because they forget to insert the newlines back between the splits when giving you the reference (red squiggles).
    2. The reference is contiguous. I haven’t tried more complicated questions so can’t rule out that they would do some kind of subquestion decomposition, but the queries I tried all just gave a single reference.
    3. The number of chunks returned is variable. This indicates some kind of “small to big” strategy where they retrieve more context than they match. Perhaps something similar to our AutoMergingRetriever https://docs.llamaindex.ai/en/stable/examples/retrievers/auto_merging_retriever.html w/ some kind of similarity cutoff but worth a bit more investigation.
    4. They have some trouble handling UTF-8.

    In general, very bleeding edge still, but look forward to seeing how this evolves.

    → 10:40 AM, Nov 7
  • This is an important slide from one of the breakout sessions at yesterday’s DevDay. But without the greater context of the presentation OpenAI’s retrieval is a black box.

    So many new toys and not sure which one to play with first!

    → 10:38 AM, Nov 7
  • Alexander Wang, CEO of ScaleAI:

    more has happened in the last year of AI than the prior 10

    I can say the same about my own life (and business!)

    we are unmistakably in the fiery takeoff of the most important technology of the rest of our lives

    It certainly feels that way. Like a whole ‘nother life I get to play. 😃

    → 10:33 AM, Nov 7
  • The amount of people proclaiming the “bloodbath” fallout from OpenAI’s release day today…it’s so telling who is insecure about their own value (and therefore just projecting that out into the world). Of course, they would only know that if they knew how narrative works. 😄

    → 12:48 AM, Nov 7
  • Overheard on Twitter:

    The real thing OpenAI did is make it way easier to be a micro-entrepreneur. A lot of companies that anyone can start and make a good living off of that don’t need venture money now. It’s actually quite delightful.

    Facts.

    → 7:14 PM, Nov 6
  • Ok. So we’re back to that moment in time when GPT-4 came out, and I would just sit stunned at the output delivered by my own app after hooking up a new model. There’s so much detail and nuance, it almost leaves one paralyzed with wonder. How does this continue? (knowing it did from 7 months ago)

    → 6:53 PM, Nov 6
  • Probably going to have to hold off Custom Models for now. 😄

    → 4:50 PM, Nov 6
  • Unbelievable presentation from OpenAI today. Assistant API basically cleans up and makes Muse WAY more easier to manage. DALL-E3 and TTS (Text-to-speech) were nice surprises as well. Going to be very busy for the next couple of months!

    → 1:22 PM, Nov 6
  • Showered, booted up, strapped in, and ready for the next wave of The Exponential Age, courtesy of OpenAI.

    → 10:46 AM, Nov 6
  • It’s the weirdest thing. The Discord server my app/niche is essentially dead—yet, my app has never been more profitable than it is right now. Usage up/subs up. Everyone creating in collaboration with my the AI I created. Strange, but lovely phenomenon.

    → 12:18 AM, Nov 6
  • Was venturing back and forth on Twitter/X but the whole Grok thing is so embarrassing I just can’t anymore.

    → 4:01 PM, Nov 5
  • The Exponential Age.

    I kinda like the sound of that.

    → 11:22 PM, Nov 4
  • OpenAI DevDay Leaks

    The leaks coming out today for Monday’s event are sending my creative builder mind reeling. I’m already seeing faster GPT-4 through APIs (had to check to make sure it was on!), but the Workplace/Teams option is the most encouraging. Already doing this manually at home for 10 months now, having a small team of AI workers to help develop my app, write content, answer emails?? We are going to be capable of doing so much more so very soon.

    OpenAI DevDay rumors:

    More than 90% rumors: - Gizmo announcement. - Features: - * Sandbox - Provides an environment to import, test, and modify existing chatbots. - * Custom Actions - Define additional functionality for your chatbot using OpenAPI specifications - * Knowledge Files - attach additional files that your chatbot can reference - * Tools - Provides basic tools for web browsing, image creation, etc. - * Analytics - View and analyze chatbot usage data - * Drafts - Save and share drafts of chatbots you’re creating - * Publishing - publicly distribute your completed chatbot - * Sharing - Set up and manage chatbot sharing - * Marketplace - Search and deploy chatbots created by other users - Magic creator / magic maker: - * Define your chatbot with an interactive interface. - * Identify user intent and create chatbots - * Live test the created chatbot - * Modify chatbot behavior through iterative conversations - * Share and deploy chatbots - Announcing Workspace and Team Plan - Team Plan: - * There are two types of Team Plans: Flexible plan and Annual plan. - * Flexible plan is priced at $30 per month and Annual plan is priced at $25 per year - * Both plans start at $90 per month with a minimum of 3 users, - * Team plan adds benefits such as unlimited high-speed GPT-4 access, 4x longer contexts, and unlimited data analytics - Support for internally shareable chat templates - WorkSpace: - * Personal/Workspace separation. - * Ability to update Workspace meta data such as role, department, etc. - * Ability to set industry/size/role/department - * Roles include team_member/manager/executive/director/business_owner/freelancer - * Business type is categorized as marketing/engineering/education_professional/design/administrative/partnerships/partnerships/research/sales/legal/other - * Job roles include developer/designer/planner - * Departments are categorized as R&D/Marketing/HR/Finance - * External system (ERP, CRM) integration available Rumor 70% or more: - Gpt4 api price reduction - GPT4 speed improvement 10~20 times (GPT4-TURBO?) - Token 32K General access Rumor 50% or more: - Dali3 API announcement Rumor 30% or more: - Image embedding - Gpt3 Open Source Release

    → 12:51 PM, Nov 4
  • Just for fun, ran the dialogue from the trailer for “Echo” into Subtxt Muse, and it correctly identified the sources of conflict in each Throughline!! 🤯 2024 gonna be crrrrazzzzy!

    → 2:33 PM, Nov 3
  • Easily the most violent “You and I” sequencing yet! 😅 👉 Echo trailer. Always looking to update the montage of Main Character/Obstacle Characters, but this one might not be for all audiences!

    → 2:28 PM, Nov 3
  • In anticipation of what I assume will be a whirlwind of a day on Monday, I’m tuning up Subtxt Muse, teaching it more narrative theory which in turn can help storytellers all across the world fully realize complete and engaging stories. I wrote about it here: Crafting your Narrative with Subtxt Muse

    → 1:51 PM, Nov 3
  • I guess I just don’t understand why, when there are so many great French and German actors, that you would hire Mark Ruffalo and Hugh Laurie. The casting in All The Light We Cannot See makes the series unwatchable—which is a shame, given how timely the novel’s themes are in our current time period.

    → 11:59 PM, Nov 2
  • Bluesky is so virulently anti-AI that I can’t completely detach from Twitter. So frustrating to see so many artists with their heads in the sand.

    Case in point: this paper post from Molick:

    The results imply that ideation and likely filtering are necessary skills in the text-to-image process, thus giving rise to “generative synesthesia” - the harmonious blending of human senses and AI mechanics to discover new creative workflow.

    I know this to be true. Both feedback from users and my own experience shifting from 2D to 3D animation back in the day. 🎨

    → 10:09 AM, Oct 31
  • I don’t use LangChain, preferring to roll my own architecture instead. But these new templates are opening up some great ideas for Subtxt and Subtxt Muse.

    → 10:04 AM, Oct 31
  • Sam Altman on AI:

    The fundamental thing about these models is not that they memorize a lot of data.

    I so wish everyone who claims “plagiarism” understood this. If they did, they would finally understand the insane levels of creativity this technology can unlock for the individual.

    The thing people like about the GPT model is not fundamentally that it has or that it knows particular knowledge, there better ways to find that. It’s that it has this larval reasoning capacity and that’s going to get better and better. That’s what this is going to be about.

    OpenAI Interview with WSJ

    → 11:09 PM, Oct 30
  • Not Chandler!!!!!!!! 😢

    → 5:52 PM, Oct 28
  • The new iOS app for Chat is so nice. So inspirational!

    → 9:25 PM, Oct 27
  • The misguided AI storytelling tweets are coming back (after about a week or so off), which always gives me a chance to write about Subtxt: The Anti-Subtxt Approach: AI, Rambling, and Missed Opportunities in Storytelling

    → 5:35 PM, Oct 27
  • Man, this gives so me so many crazy ideas. 🤓

    → 11:12 PM, Oct 26
  • Basically just waiting for Nov. 6 to get my new AI marching orders. 🫡

    → 2:05 PM, Oct 26
  • Next quotable quote: “AI with people will replace people without AI. 🤗”

    → 11:31 AM, Oct 25
  • “Adapt or face extinction.” Yep. This, from Marily Nika, currently speaking on AI in 2024 (she’s an AI Product Lead at Meta). Love the no-pulling punches approach.

    → 11:28 AM, Oct 25
  • When your own invention roasts you better than any median human could. 😂

    → 3:53 PM, Oct 22
  • Oh, that’s right. New Blink. 🎸🥁

    → 4:14 PM, Oct 20
  • Maven reviewed my cohort and made suggestions. Followed most of them (all of them), and the new re-branded Write Great Stories with AI is live! After two years and 11 cohorts, looking forward to shaking things up. 🚀

    → 2:02 PM, Oct 20
  • That point when you’re so excited about your work you have no time to blog or create content.

    → 7:57 PM, Oct 18
  • Nothing quite like a live demo to set you straight and give you a clear direction as to where to take things next. Never underestimate the power of user confusion! Introducing Subtxt’s New Feature: Narrative Agents

    → 3:50 PM, Oct 16
  • After the big demo to a live audience on Friday night, I had a ton of direction for Subtxt. It also just so happened to sync up perfectly with iA’s latest piece on preparing for NaNoWriMo: Elevate your Nanowrimo and Daily Writing Game with Subtxt

    → 7:00 PM, Oct 15
  • Mornings now are so much more peaceful. I used to think the photo icon for Bluesky was so cheesy, but now is so apropos given the shift in social media over the past week. And I’m starting to notice that more of the people I’m interested in following are doing the same. So happy ☀️

    → 12:18 PM, Oct 15
  • New Welcome page for Subtxt users. Can’t wait to roll it out tomorrow. 🤗

    → 1:54 AM, Oct 15
  • This is crazy. (Full disclosure: i’ve never cleaned it 😊)

    → 3:48 PM, Oct 14
  • Finally read the NYMag hit-piece on Altman and OpenAI. Hoo boy. A lot of misconceptions here on AI and the creative process.

    → 3:15 PM, Oct 14
  • How is this even possible?!

    → 12:52 PM, Oct 14
  • The sun is going out!! ☀️🌑

    → 9:18 AM, Oct 14
  • Whew,😅 Subtxt first live demo in the bag. SO MUCH FUN demoing what I’ve been working on these past years to real live people. Only had enough time to go over 5% of what the app can do, but so much clearer on the direction to take the next step.

    → 2:11 AM, Oct 14
  • Off to give a live presentation of Subtxt. 😃

    → 2:59 PM, Oct 13
  • Oh, you know. Just having a discussion with AI about how best to prepare my pour-over coffee.

    → 8:17 PM, Oct 12
  • Whew, what a day! 😅 Teaching in the morning, smashing bugs in the afternoon, and penning a brand new blog post on Romantic Dramas and Story Structure–all while preparing for a pretty big presentation tomorrow night.

    → 7:36 PM, Oct 12
  • Happy to see Bluesky start to come to life in the last couple of days.

    → 2:26 PM, Oct 12
  • Well. Here we go.

    → 4:09 PM, Oct 11
  • I usually can’t stand this guy, but DHH sure gives good conference.

    → 3:55 PM, Oct 11
  • So, even though I’m entering another decade, we’re just gonna keep having these nightmares where I wake up panicked that I don’t have enough credits to graduate, right?

    → 12:59 PM, Oct 11
  • 2023 problems.

    → 8:21 PM, Oct 10
  • Need to watch “her” again so I can figure out what to talk about with ChatGPT-Voice. ⭕️🤖❤️

    → 7:59 PM, Oct 10
  • So basically, in about 20 minutes, I improved the output of Muse while eliminating alot of redundancies that were wasteful in terms of resources. NO WAY I would have figured this out on my own. This is “orders of magnitude” great! 😂

    → 6:50 PM, Oct 10
  • As a one-man team, I greatly appreciate ChatGPT’s ability to analyze my app’s usage and make intelligent suggestions on how to increase improve the experience all around. Really feel like I have employees helping to explain things out for me.

    → 6:32 PM, Oct 10
  • The latest update to Subtxt takes storytelling AI to the next level by nailing nuanced character roles like never before. Say goodbye to clichés and hello to richer stories. 📖

    → 4:42 PM, Oct 10
  • I mean, this is just absolutely incredible. DALL-E 3 in Chat is amazing!! (#4 perfectly captures my work day!)

    → 2:19 PM, Oct 10
  • I nominate “orders of magnitude” for the most overused phrase of 2023.

    → 10:04 AM, Oct 10
  • First fine-tuning using OpenAI’s new API turned out way better than I had expected. Response was clearly defined (valid JSON), and was much better in terms of alignment with narrative concepts. Very excited to do more. 😃

    → 8:11 PM, Oct 9
  • Checking out the opening commencements of the first-ever AI-Engineers Summit conference. A welcome distraction for today.

    → 2:54 PM, Oct 9
  • Saw Past Lives over the weekend. Totally get what was going on, BUT…kinda wished there was a bit more “story” in there so that I could make sense of the narrative…Analysis of Past Lives

    → 1:42 PM, Oct 9
  • The monthly Dramatic Users Group analysis is on for tomorrow night, Tuesday Oct. 10 @ 7pm Pacific. Join theory co-creator Chris Huntley (and me!) as we take a look at Men In Black through the eyes of Dramatica. Event details in the Discord.

    → 12:36 PM, Oct 9
  • Finally getting a chance to take this whole ai-generated storytelling thing to the next level…🚀💪

    → 8:42 PM, Oct 8
  • With doom-scrolling a thing of the past, I’m finding more time to read blogs. As a result, I’m finding it easier to respond with more thoughtful pieces that get across my own unique viewpoint when it comes to AI and creativity. AI, Copyright, and the Narrative of Ownership

    → 6:47 PM, Oct 8
  • Early last week I told my wife that I was finally quitting Twitter/X cold-turkey, that I no longer wanted to subject myself to the latest in AI tech and MMRs interspersed with the worst part of humanity. My reasoning was that the next year would be one of insanity.

    Looks like I quit just in time.

    → 5:40 PM, Oct 8
  • Found yet another AI writing app built on push-button storytelling on YouTube last night. Great opportunity to explain the Subtxt difference: The Push-Button Paradox: Why True Storytelling Demands More

    → 3:39 PM, Oct 7
  • On a great roll with publishing pretty much every day on Narrative First. Feels like old times, just every day instead of every week! 😅 Today was based on some questions from the last cohort session: Navigating the Relationship Story Throughline of a Complete Story.

    → 1:51 PM, Oct 6
  • Elon sure makes it easy to walk away from 14 years of Twitter. I jumped off early in August, but then came back thinking it helped with my business. The latest, no-link titles and now no re-tweet/like buttons, and then just him 😂 at a serious question about how engagement and the algorithm has deteriorated in the last couple of weeks, really was it for me.

    If anything, it does give me a moment to think about how I could be better for the users of my app. More attentive, and more conscientious of feedback.

    → 12:12 PM, Oct 6
  • The robot apocalypse is upon us! This is what the movie The Creator is warning us all about: using AI to create a superficial meaningless world. 😲 (Not really, but pairing ChatGPT with Pixar’s “22 Rules of Storytelling” will get you the same results).

    → 3:21 PM, Oct 5
  • New double-tap on the iOS watch is fun! 😁 Message comes in, double-tap hands-free to open up dictation, speak back your reply, double-tap to send. ✅ 📨

    → 3:49 PM, Oct 4
  • This past weekend I took a chance on The Blackening. While funny (the trailer spoils most of the jokes), the narrative lacked emotional cohesion, which would have really made it more memorable. Even on the most simplest of projects, you can strive to do more, and deliver a complete story.

    → 3:40 PM, Oct 4
  • Welcome to developing an app with OpenAI’s API, where you never ever get any rest! 😅 (Fine-tuning Function Calls just rolled out quietly … which means Subtxt and Subtxt Muse are going to get even better).

    → 6:16 PM, Oct 3
  • Over the weekend, Muse passed 100,000 questions answered. That means since Jan. 27, Muse has been answering, on average, 400 questions a day, or one question every 3 minutes. 💪😊

    My intention when I set out to build Subtxt back in 2017 was to find a way to exponentially increase my ability to help writers get closer to the story they wanted to tell.

    I had no idea it would get to this point. 🙏

    → 4:03 PM, Oct 3
  • I can highly recommend Zimmer’s latest soundtrack for The Creator (had no idea he did the music for it). Of particular note is this track: True Love. Perfect for afternoon sessions of creativity. 🎬 🎵

    → 2:37 PM, Oct 3
  • Just dropped a fresh analysis on Christopher Nolan’s Dunkirk. Ever wonder how three separate timelines can lead to one unified emotional punch? Subtxt and Dramatica theory have you covered. Stay tuned! 🎬📚

    → 2:28 PM, Oct 3
  • OK. It’s time to just release GPT-4V. I’ve seen enough to know I’m ready for it.

    → 2:00 PM, Oct 3
  • So thankful for this feature on my Sonos system. I can finally fix the lip sync issues in Across the Spider-verse!

    → 5:14 PM, Oct 2
  • I received a few last-minute requests to join tomorrow’s cohort—the last of the year. I’ve extended the payment deadline to Thursday, so yes, you can still hop on board. Let’s build better stories together!

    → 2:01 PM, Oct 2
  • Wrapping up this week with a short little blog post on Narrative First about the difference between storytelling and story. This is how Subtxt works: it models the underlying structure and inherent meaning of the text, not the text itself. AI currently models the words, so thats why most of the stories it creates are horrible.

    In short, if Disney trains its AI on Ashoka or The Mandalorian writers worth their salt have nothing to worry about.

    → 3:13 PM, Oct 1
  • AI will end the careers of those writers who believe that AI will end their careers.

    → 10:35 AM, Oct 1
  • While missing an invite to OpenAI’s DevDay was initially disheartening, a glance at the schedule eased my FOMO. It seems I won’t be missing much; the event doesn’t offer many groundbreaking insights this time around. In contrast, the upcoming AI Engineers Summit seems far more enticing with a jam-packed agenda promising depth and breadth in discussions.

    That said, what does interest me is that be line about strategies for probabilistic programming over deterministic approaches. That’s been the hardest, yet most exciting, aspect of this new shift. There’s so much potential–but sometimes you have to see the potential in order to know what to do with it.

    → 12:46 PM, Sep 30
  • AI enthusiast champions new algorithms to obliterate the in-betweening process:

    Animation lovers, rejoice! The labor of love that is hand-drawn 2D animation is about to be revolutionized by a pioneering deep learning technique that automates the painstaking inbetweening process.

    As someone who worked as a rough in-betweener for Disney in the mid 90s, I can tell you that there was no part of it that was painstaking. It was a beautiful experience and I loved every minute of it.

    This is a revolution that’s kind of missing the point. 🎨

    → 1:29 AM, Sep 30
  • Man, what a bummer. Just got my reject notice for OpenAI’s DevDay. Was so looking forward to it. I guess already having a profitable app running in part off their API disqualified me. ☹️

    → 5:52 PM, Sep 29
  • The final Build a Better Story cohort of 2023 kicks off next Tuesday. Still a few seats left for those looking to elevate their narrative game! Dive deep into story theory with like-minded individuals and end the year on a high-note. 🚀

    → 11:47 AM, Sep 29
  • After this week, I think I’m up to 50x.

    → 7:58 PM, Sep 28
  • Just rolled out Inline Editing of Questions within Muse, and couldn’t be more excited. I know, it’s something that’s already there in ChatGPT, but there are some changes I’ve made to be more apropos for storytellers. The technical challenge on this was really fun to work through this week. Love how it all turned out.

    → 7:32 PM, Sep 28
  • Another week, another Arxiv paper where ML researchers attempt to write stories with LLMs…without knowing how stories work. If only there was a predictive narrative framework already in production…

    → 12:09 PM, Sep 28
  • I’m really really OK with this. 🤗 Love how it gives a bit of context, yet doesn’t just grab everything from the page. Just a gentle encouragement to link the original source.

    → 3:20 PM, Sep 27
  • Taken me all of ten months to finally figure out what direction to go in with all these AI upgrades. Ten months to realize that I was already on the path, just the horizon had changed. The Birth of Narrative First and Subtxt

    → 11:08 AM, Sep 27
  • Matt Wolfe is great, and his latest on AI hype: copyright and authors suing is wonderfully on point. The Creative Table of AI, Authors, and Creativity.

    → 1:18 PM, Sep 26
  • After a 5-month WGA strike and a tentative agreement finally reached, today’s blog post on the ‘Reversal Curse’ in language models couldn’t be more timely. It digs into why AI can’t capture the magic that makes human storytellers truly irreplaceable.

    → 12:31 PM, Sep 25
  • Can we all just agree that Sky is supposed to be Scarlett Johanssen and Breeze is supposed to be Tom Holland??! (talking about the new OpenAI voices).

    → 12:01 PM, Sep 25
  • What a W for optimism today: the WGA strike and AIs that can speak, see, and hear. No idea how you could not be bullish on the next twenty years.

    → 11:54 AM, Sep 25
  • Paralyzed with excitement now that the film “her” is reality.

    → 10:07 AM, Sep 25
  • So, NeMO Guardrails is essentially OpenAI’s Function calls but semantically-aware of the questions themselves and where to direct the response? And I can model the flow of questions asked? 🤯

    → 6:22 PM, Sep 24
  • That Follower/Following ratio is probably a good lesson for everyone looking to change the world.

    → 1:08 PM, Sep 24
  • Shifting from a traditional corporate role to forge your own path as a solopreneur disrupts the “standard” narrative of success, causing some to question or even disparage the choice. It’s like you’ve gone off-script in a well-known story, creating a bit of cognitive dissonance for those who haven’t considered alternate routes to fulfillment.

    The classic retort of “Oh yeah? Then do it!” perfectly encapsulates the challenge and reward of choosing a non-traditional career. It serves both as a challenge to those questioning your choices and as a reminder that crafting your own life story requires its own set of rules, structures, and hard work—just like any good narrative. Cheers to those willing to write their own compelling stories!

    → 11:02 AM, Sep 24
  • Isn’t it obvious that the Action button goes straight to ChatGPT?!

    → 10:17 AM, Sep 23
  • LOVE the weight and feel of the natural titanium. 😊

    → 6:24 PM, Sep 22
  • The new OpenAI Cookbook is SO nice, and so badly needed. The previous was a nightmare to navigate through, this update is quite elegant and should give everyone an easy in.

    → 3:49 PM, Sep 22
  • When Technology Can Write, Will Storytelling Suffer?

    Hey folks, Jim Hull here, and let’s talk about something that caught my eye this week—the release of DALL-E 3, which generates images from text. while the demo was captivating, what struck me was the notion that the “story” it created at the end could be considered a real story. As someone deeply immersed in story theory and structure, it raises a big red flag for me. If you thought explaining the essence of good storytelling to Hollywood execs was challenging, buckle up because it’s about to get a whole lot trickier.

    Complete Sentences Don’t Make a Story

    I mean, look, DALL-E 3’s abilities are amazing, but just because it can string sentences together doesn’t mean it’s crafting stories with meaning. Storytelling isn’t just about creating a series of events or dialogues; it’s about thematic richness, character arcs, and those gut-punch moments that resonate with audiences. It’s about weaving throughlines and subtext, something a machine just can’t get… at least not yet.

    A Flood of Content, but Not Substance

    Now, what happens when every aspiring storyteller gets their hands on DALL-E 3? We’re gonna be deluged with content that may look like a story and sound like a story, but lacks the soul of a story. Think “Ashoka” was a storytelling misstep? Imagine a world flooded with similar narratives, only worse because they’re mass-produced by next-word algorithms.

    The Real Challenge: Educating the Masses

    What’s concerning here is the scale. Teaching the intricacies of storytelling to a room full of studio execs is one thing. But how do you educate millions of people who think they’re telling meaningful stories just because they have a machine that writes in complete sentences? That’s a daunting task, but it also means there’s an even bigger market for folks who understand the mechanics of effective storytelling.

    A Call to Action for Story Theorists

    This is a moment where the teachings of story theory and structure become more critical than ever. It’s our responsibility to lead the way in educating people about what makes a story compelling and resonant. Technology is a tool, not a substitute for the human elements that give a story its heart.

    So, get ready for a bumpy but exciting ride ahead. There’s a lot of work to do, and I’m up for the challenge. How about you?

    Catch you later!

    → 12:52 AM, Sep 22
  • The first episode of One Piece has more story than this entire season of Ashoka. 😄

    → 1:04 PM, Sep 21
  • OpenAI sneak previews DALL-E 3 which looks fantastic (closer to Midjourney at least), BUT without an API for non-Enterprisers like me I’ll sadly have to wait! 😩

    → 10:53 AM, Sep 20
  • First time working with Data Analysis within ChatGPT. 😳 Looking over the past two years of students to find commonalities of interests, and oh my … this saves so much time.

    → 11:57 AM, Sep 19
  • Super cool. Some API requests to OpenAI failed, and for some reason my response consisted of Python debugging code showing how it teaches itself how to fix mistakes (hallucinations, etc.). Hour later, same request was golden (no error).

    → 8:46 PM, Sep 18
  • Kicking off The World of Narrative Course this week in the Narrative First School.

    Ready to decode the magic of storytelling? Join us this Thursday as we kick off our exciting new course, The World of Narrative! Open to all students, we’ll dive into the groundbreaking Storymind model to unravel what makes narratives tick.

    First class covers requirements & expectations. Don’t miss this chance to level up your understanding of story theory and structure! 🚀

    Looking forward to making this class–which I’ve taught for years now–accessible to more people.

    → 3:42 PM, Sep 18
  • Alot to keep up with today: new interactive widgets across all my productivity apps (Things, Timery). OpenAI’s DevDay, and now the release of gpt-3.5-turbo-instruct. Got lots to today!

    → 2:44 PM, Sep 18
  • Fingers crossed!

    → 12:30 PM, Sep 18
  • Another year, another new iPhone. 😊 Grateful once again.

    → 5:09 AM, Sep 15
  • Once a year I have to get up at 4:30 in the morning. Cmon Apple. I do the same thing every year. Can’t you just automatically send me the upgrade without the sleep pattern disturbance?!

    → 8:16 PM, Sep 14
  • Just finished the TENTH Cohort of the Build a Better Story class. 😅 Really love teaching this material and very thankful there are so many interested in it as well.

    → 12:56 PM, Sep 14
  • Published a walkthrough of how to interact with Subtxt Muse to help you “storyform” your next great story. It’s essentially a process of finding the meaning of your story by working with the AI to your story’s deep thematic meaning. How To Storyform with Subtxt Muse

    → 6:05 PM, Sep 13
  • Just learned about Embedded Growth Obligations (ego), which gives me permission to slow the heck down. 😴

    → 12:49 AM, Sep 13
  • Going to get back in the habit of recording more YouTube videos teaching How Stories Work while also showing off Muse. Today’s 10-minute video is all about The Four Throughlines of a complete story.

    → 6:32 PM, Sep 11
  • We’ll be taking a look at del Toro’s Pinnochio tomorrow night at the monthly Dramatica Users Group meeting. All are welcome.

    → 4:47 PM, Sep 11
  • Best part about being home.

    → 4:13 PM, Sep 10
  • Paso paradise.

    → 6:21 PM, Sep 8
  • Just read this on Twitter/X and it makes perfect sense. That once it starts hallucinating and making mistakes, it continues to model those mistakes more and more. 🤯 Going to add an edit button next week to Muse…x.com/yacinemtb…

    → 4:42 PM, Sep 8
  • Just published a new page in the Guide about going the extra distance when using AI to develop a story, and how its necessary to look beyond the superficial next-word prediction to get down to the Sources of Conflict present in every aspect. Why Is This a Problem?

    Sub-header should be: “Why push-button storytelling is not enough…” 💪

    → 3:06 PM, Sep 7
  • Guess I know what I’m doing in November 👉First ever OpenAI DevDay 🤓

    → 10:47 AM, Sep 6
  • First student just applied for October’s cohort and I haven’t even started marketing it yet. 😃

    → 2:00 PM, Sep 5
  • My current go-to prompt for anything tldr related: “write me a pithy subheader (7-8 words, titlecase, no “unraveling” or “unleashing”)“. Works every time. 💪

    → 1:43 PM, Sep 5
  • GPT-4, not sure how I would do it without you. User writes in and is confused about pricing tiers. I jot down quick thoughts as to the difference, have GPT-4 expand that into a friendly support reply. I then ask it to rewrite its answer as if documentation (for inclusion into the Guide. And then finally, have it take its explanation of the differences and turn it into bullet points for use on the landing page of Subtxt.

    It’s been eight months now, and I’m still not used to all I can get done in a single day.

    → 3:39 PM, Sep 4
  • Erik Grönwall is my latest and greatest discovery. Perfect soundtrack for coding and writing.

    → 5:45 PM, Sep 2
  • Predicting the Story of The Creator

    In a first-ever event for the Subtxt and Dramatica theroy, the complete Storyform for a film that has yet to be released is now in Subtxt! 😃 Just based on the excellent trailer, which reveals all Four Throughlines in just 2 minutes as well as the Pivotal Elements of Control and Free (Joshua is Control, Maya is Free), one can assess the entire thematic construct of the film.

    What does this mean?

    It means one can predict what will be happening in each and every Act, just based on the meaning of the film presented in the trailer. 🤯

    → 4:34 PM, Sep 2
  • Added the Storyform for Vengeance into Subtxt: Vengeance

    → 4:00 PM, Sep 2
  • My eyes are very thankful that I finally switched PHPStorm over to Drasner’s Night Owl theme (that I’ve been using in VSCode for years). Highly recommend!

    → 3:17 PM, Sep 2
  • Finally realized this week that users would much rather know exactly what is going on (“Network error”) rather than a story-driven, character-based response (“I’m sorry. it appears as if I’m having a brain freeze at the moment.”). Most chalked up the latter to something wrong with the app, everyone appreciates the former as a problem with outside services.

    It’s pretty much exactly what ChatGPT provides, so it’s not ground-breaking. But apparently, the more obvious and simple the more forgiving users are.

    → 5:52 PM, Aug 31
  • The Art of Crafting Stories in an AI-Driven World. Turns out the answer is being able to ask the right questions.

    → 1:38 PM, Aug 31
  • Being a solo-preneur can be quite lonely sometimes. That’s why I’m much appreciative of Disparu and Nerdrotic, as watching their takedowns of Ashoka reminds me of what it used to be like hanging out at lunch with friends at Dreamworks.

    → 2:18 PM, Aug 30
  • Recorded a much-needed 15 min. walkthrough of Subtxt for new users. Until the AI can walk you through the process from beginning to end, this is the approach most should use. Gotta say–the new Regenerate feature in Descript is nutso - works great!!

    → 2:22 PM, Aug 25
  • Working within the emergent properties of machine learning and large-language models, I’m always constantly amazed at how a giant leap forward will be taken, all without my doing (I mean, it is my doing, but I didn’t expect it).

    Case in point, I’m working on making it possible for writers to develop their entire stories just in chat–currently starting out with just creating a new story and titling it (something that used to be handled by your traditional forms).

    For fun, I always try to come up with some new story ideas just to see what would happen. As you can see in this, I didn’t expect Muse to give me back the entire narrative structure! I even asked it for a name before I realized it already did that in the background.

    Simply astounding.

    → 8:15 PM, Aug 22
  • Need that Earthquake alert about 10 seconds earlier. I was already under the dining room table!

    → 2:44 PM, Aug 20
  • Watched Twister tonight, so I’m all ready for tomorrow. “Finger of God!”

    → 12:34 AM, Aug 20
  • Man, I love being able to work with seriously talented storytellers. Especially on Saturdays! 😃

    → 5:18 PM, Aug 19
  • There’s probably nothing cooler than a native-speaking Parisian using GPT’s innate ability to effortlessly translate my English concepts of narrative structure into French and back again. And more importantly, for them to be overly enthusiastic about the results. 😃 🇫🇷

    → 10:18 AM, Aug 18
  • Pretty sure I just unlocked an unplanned, but potentially amazing, feature of Muse: Cinema-therapy. Enter something you’re struggling with now in your real life and plug in to someone else’s narrative, and perhaps find a meaningful way to move on from what is holding you back.

    → 2:26 PM, Aug 16
  • Published a blog post about using Subtxt Muse to help develop a story, by leaning on the narrative structure of one of the greats: Stephen King’s Stand By Me (and somewhat influenced by my recent foray into The Blade Itself 😊. I admit, it helps to know what to ask, but when you do the story development process becomes really fun.

    → 6:36 PM, Aug 15
  • It will never not blow me away that this is even possible. I will be Function Calling everything in Subtxt.

    → 7:38 PM, Aug 14
  • Watching the sun rise and the changing colors of the boulders in Joshua Tree National Park yesterday (after a night of 60+ meteors an hour!). Truly a beautiful experience.

    → 9:01 AM, Aug 14
  • Why can’t I stop watch and re-watching the Nueva York Train Chase from Spider-verse? I’ve probably seen it eight times already, just on a constant loop! Such a perfect blend of character, action, cinematography, animation, dialogue, music and editing. 🚄💥

    → 9:54 AM, Aug 11
  • Pleased to announce this latest feature for Subtxt: Muse Personas. Guide and instruct Muse to be the kind of super-intelligence you’d like to collaborate with. Create your own virtual writer’s room!

    → 1:22 PM, Aug 9
  • The recording from last night’s Dramatica Users Group analysis of Mary Poppins is now up in the Narrative First School. Glad to finally know why it’s such a difficult (i.e. boring) film to sit through - no story! (or, at least, two incomplete malformed stories).

    → 10:10 AM, Aug 9
  • Found this while cleaning up Narrative First. Calvin & Hobbes has always been an immense inspiration to me. I had no idea this is where I got the idea for Subtxt Muse. 😄

    → 5:01 PM, Aug 8
  • New cohort underway! 🚢 Great group of writers and even some non-storytellers in the mix. I was happy to finally be able to find a place to use this slide as well. Looking forward to the next six weeks.

    → 1:08 PM, Aug 8
  • Feeling good about this one today: got all 19 episodes of “The Page Master Course” - my first cohort ever, loaded up into the Narrative First School. Something like 40+ hours of deep-diving narrative theory.

    → 8:57 PM, Aug 7
  • In Subtxt, we call this kind of narrative “Timespace”, in that the organizing principle of the narrative is time (in contrast to Spacetime).

    → 9:19 AM, Aug 6
  • Finally finished Life’s Work by David Milch 📚. So crazy to read how everything unraveled for him, but delighted to be inspired by his words:

    The process of storytelling, when it’s wrung from the heart, is a process of discovery, and self-discovery, and that has remained true with this work at this time.

    I had a feeling the Deadwood movie was all about him, and so heartwarming to hear how everyone came together to complete it.

    If you love a good story, and are fascinated by the process of putting one together, I highly recommend it.

    I’m making the decision to believe that this process of storytelling, begun again, will have a formed and happy result. That may or may not turn out to be an illusion, but one at least has the solace of knowing one entered with a willing heart.

    → 9:55 AM, Aug 5
  • Just rolled out virtual Personas for Muse. Choose from which virtual “personality” you want to help you collaborate on your next story. Build your own virtual writer’s room! 🛸🤖🚀

    → 9:26 PM, Aug 4
  • Added BlackBerry into the list as well. Really fun film. And well told.

    → 12:53 PM, Aug 4
  • Always loved this movie. Found some time today to add it into Subtxt.

    → 12:30 PM, Aug 4
  • Loved Blackberry. Of course, can’t waste the opportunity to upload its narrative structure into Subtxt, only to find it wasn’t one of the 4000+ already in there. My favorite is that the model predicts, based on this premise, that Mike would be faced with issues of the Future during the last act. No better way to interpret this then the arrival of the iPhone.

    → 10:54 AM, Aug 4
  • Added Calendar Events to the new Narrative First School. Love that there is a one-stop shop now for everything great-storytelling related! 😊🚀

    → 7:46 PM, Aug 3
  • Join Dramatica co-creator Chris Huntley as we dive into the narrative structure of Disney’s “Mary Poppins” using the Dramatica theory of story for context.

    → 5:38 PM, Aug 3
  • Incredibly, Subtxt has been featured in the New York Times! To say I’m honored is an understatement. The fact that they explained what it does so clearly and pleasantly is the real gift. 🙏🏼

    Among the start-ups are Stockimg, which can produce book covers; Storywizard, a program that creates children’s stories; Subtxt, which acts as a writing coach by helping authors expand on a concept or develop characters; and Laika, an A.I. program that claims to mimic the prose of writers like Jane Austen and Edgar Allan Poe.

    → 9:56 AM, Aug 2
  • Currently reading: Life’s Work by David Milch 📚(and adding Epilogue to my home screen 😊)

    → 6:49 PM, Aug 1
  • David Mitch on “writing” Deadwood:

    All storytelling is collaborative in one way or another. Nobody writes an episode independently. Once the scene is generated, we don’t let it ripen particularly, usually the scene was written either the day it’s shot or the day before. No finished script before the shooting begins. If the shoot was an average of twelve days, the script was finished on the twelfth day of the shoot. The idea of being an auteur, I don’t believe in that. I don’t care whose name is on a script. We are organs of a larger organism which knows us although we do not know it. I regard myself as a vessel of whatever that larger organism is, its instrument, rather than as the source of the scenes.

    I love the romanticism of this approach, and just makes me love and appreciate the series all the more.

    → 9:42 AM, Aug 1
  • Unbelievable. From nothing to full-blown web app integrated with my decades old blog in just 5 days! 🤯 The new Laravel/Folio/Volt method is such a delight. ⌨️

    → 4:27 PM, Jul 31
  • I did NOT know Paul Reubens went to CalArts! 🎬🚌

    I always sort of felt like Pee-wee Herman was performance art a little bit, and what I personally liked about it was that I was the only per­son who knew that. I was in art school at CalArts [California Institute of the Arts] just prior to joining The Groundlings and developing Pee-wee Herman. I almost felt like Pee-wee Herman was conceptual art because I went to great lengths to make people think that he was a real person.

    → 2:03 PM, Jul 31
  • Somebody just told me that the new slogan for Twitter is “Blaze your glory.” I mean, if there’s any indication someone is stuck in the early 90s. 🙄

    → 9:25 AM, Jul 31
  • It is ridiculous how quickly I’ve been able to build a new web application using Laravel Folio, Laravel Volt, and GPT-4. Started Thursday, ready to roll out later today. Complete with full auth, gated sections, and beautiful integration with the blog I started 15 years ago.

    → 4:43 PM, Jul 30
  • Finally saw Drive My Car last night as it’s been my wish list for awhile. Completely sublime and takes you to a different world. 🚗 🎬

    → 11:27 AM, Jul 30
  • Added Split Conversations into Muse just now, and not sure why I didn’t think of this earlier. Writers can now brainstorm down one rabbit hole, then crawl back up and split the conversation so they can try a completely different rabbit hole. 🐇

    → 8:27 PM, Jul 28
  • This comic on designing conversations has me thinking about ways to vastly improve the experience of writing a story with Muse. Imagine threaded conversations that spill out into different dark and mysterious avenues.

    → 1:46 PM, Jul 28
  • Really wish I could have done a better job of explaining to my kids this reality of life, that life is what you make of it. It’s your reality, and you don’t have to stay within someone else’s box. Who knows…still got some time!

    → 5:04 PM, Jul 27
  • Of course, the moment I finally break free from the dumpster fire that is Twitter/X someone with a huge following posts a really nice “tweet” about Subtxt and the subscriber count shoots up. “Just when I thought I was out!!” 😅

    → 4:49 PM, Jul 27
  • The problem with developing with an AI assistant (GPT-4 for me) is that now, I can do so much in one day that I really have to find ways to take a break. It used to be that I would take a break when things get frustrating–but now, I never run into roadblocks so it just keeps going and going!

    → 11:15 AM, Jul 27
  • New year starts today!

    → 12:32 PM, Jul 26
  • Why I won’t be joining Threads:

    “For You is the default experience when you open the Threads app,” Meta spokesperson Seine Kim tells The Verge. We’ve asked Kim if Meta plans to add a way to set Following as a default in the future.

    Making an algorithmic “For You” tab the default position is toxic and manipulative. It also happens to be one of the main reasons why I had to finally delete Twitter.

    → 2:55 PM, Jul 25
  • Really can’t believe that I just deleted Twitter off my phone. It was so helpful getting my message across and helping to build my business, but I can’t justify maintaining a feed that could be changed in a whim.

    Looking forward to re-familiarizing myself with RSS and my email list. 📡📧

    → 1:55 PM, Jul 25
  • Barbie totally reminded me of Phantom Menace opening night—except that the positive vibes and enthusiasm in this one died so much quicker. I’ve never seen so much utter disappointment on the face of fans who were expecting something completely different.

    → 10:42 PM, Jul 24
  • Watching the final sequence of National Lampoon’s Vacation last night, I was amazed at how beautiful and GREEN my hometown used to be in the early 80s. Driving past Magic Mtn. this morning was depressing at how brown and decimated the landscape has become over the years.

    → 3:23 PM, Jul 24
  • AI chatbot I created helps write the best bio ever for me. 😊

    → 7:10 PM, Jul 22
  • Writing with Muse tonight. Inspired by “The Menu” again, I start building my story off of that—only to have the highly complex chatbot I created remind me I still had to “write the same thing!” 😆

    → 1:49 AM, Jul 22
  • Trolling the AI Tech World

    One of the fun things about Subtxt and Dramatica is being able to recognize story in a way that kind of sets you apart from everyone else. It’s like you know this amazing secret that gives you such a wonderful—and insightful—appreciation of the world.

    Case in point. The South Park Showrunner AI

    It’s one massive troll—a true masterpiece.

    If you go through their site, or read the white paper, you can see this completely fabricated nonsense. Nonsense likely created by Matt and Trey.

    And everyone everywhere is falling for it.

    I mean, c’mon. Just read this tweet:

    We also used it bc the creators of South Park made the brilliant deepfake company Deep Voodoo - which created the ingenious Sassy Justice deepfake TV show and deepfakes of many celebrities and actors. Probably the most important company at the intersection of creativity and AI (meaning a co that has creatives in leadership that want to make art rather than only a tool for other creatives).

    And we loved their GPT episode recently probably the best skewering of the tech so far! And the pomposity of the Valley.

    Much better than anything AI is going to do! They’re brilliant artists! Much better than us too! Geniuses!

    lol. You’re trying to tell me this isn’t Matt and Trey?? 🤣🤣

    This “Showrunner AI” is so clearly aimed at showing how ridiculous everything is getting. Even a supposed NVIDIA PhD scientist from Stanford is buying into it (check out the attached image).

    But I’m even starting to think Dr. Jim is a completely simulated AI as well! And what does this have to do with knowing Subtxt/Dramatica? Go no further than this latest tweet:

    We tried a lot of things to try to solve infinite story! We experimented with a Joseph Campbell system with archetypes and roles to create tension and conflict among the AIs.

    But ultimately even though George Lucas may be able to make Star Wars with Campbell archetypes there isn’t (in our opinion) an underlying formula or periodic table for all stories. That would mean we could all make Star Wars!

    The infinite stories we were getting from a more roles based approach didn’t lead to interesting ongoing narratives. The hero’s journey has a beginning middle end - we wanted endless story.

    There actually IS a periodic table of story elements.

    One they would know—if they were actually a part of this world! 😂

    → 5:26 PM, Jul 21
  • With today’s release of NativePHP, the desktop version of Subtxt is on it’s way. 😃

    → 12:34 PM, Jul 20
  • Guess I’m making the switch to the TALL stack. 🚀 Today’s presentation on v3 of Livewire (and I’ve been watching this develop for years now) totally sealed the deal. Should be a fun couple of months.

    → 12:31 PM, Jul 20
  • When the massively-intelligent narrative chatbot you built knows the inside joke about you…(a writer sent this to me as it was something that popped up on his device) 🤣

    → 5:30 PM, Jul 19
  • Building Hollywood's Version of ChatGPT

    Wired covers Subtxt–without knowing about Subtxt:

    First, generative AI will eventually be a valuable tool in some creative realms, potentially including script writing, but only if the AI has been built from the ground up for that task.

    That would be the app I started building six years ago.

    Second, the flaws of today’s generative AIs make them unsuitable for serious work, especially in creative fields. General purpose AIs, like ChatGPT, are trained on whatever content the creator can steal on the internet, which means their output often consists of nonsense dressed up to appear authoritative. The best they can do is imitate their training set. These AIs will never be any good at creating draft scripts—even of the most formulaic programming—unless their training set includes a giant library of Hollywood scripts.

    The first part of this is right–the second part is woefully wrong.

    Training LLMs on scripts will just result in a preponderance of badly-written, meaningless screenplays.

    Training LLMs on a narrative framework that predicts a sequence of events that align with artistic intention and meaning?

    That’s the only way forward.

    → 3:55 PM, Jul 19
  • So, I guess with Laravel Folio and Laravel Volt I can just drop Nuxt completely then…🤔

    → 2:46 PM, Jul 19
  • Six years ago this week I stumbled across Laravel in my quest to learning what would be the best way to develop my own narrative application. Best accidental discovery ever. Today’s release of Laravel Herd is further confirmation.

    → 2:06 PM, Jul 19
  • Ben Orenstein’s talk on Developer-turned-Founder was great. The usual list of mistakes, but presented clearly and beautifully. Side note: you can now tell when a talk is nearing the end based on color temperature of the background (thanks to iA Presenter!).

    → 12:14 PM, Jul 19
  • “Push-button storytelling” is compelling for those who have never really written a story. They miss the fundamental reality that compelling stories communicate something more than just the story itself. The best interface is a blank page devoid of boxes and avatars.

    → 10:36 AM, Jul 19
  • I can definitely recommend Other People’s Children—both from personal experience, and from just a great cinema experience. 🎥

    → 12:56 AM, Jul 19
  • Interested in how stories work? Just recorded a quick video on how you can use Subtxt Muse (the massively intelligent narrative chatbot) to expand your appreciation of all things story! Learning about Storyforms - How Stories Work

    → 6:18 PM, Jul 18
  • Rite of passage unlocked: finally got to see a child (step-daughter) rolling up in the driver’s seat after passing permit test. 🙌

    → 3:51 PM, Jul 18
  • The best part about today’s final cohort class? Students were able to accurately predict the entire narrative structure of the upcoming movie, The Creator…just based on a 2-minute trailer.

    🎤 🫳

    → 2:15 PM, Jul 18
  • Day one of my Twitter detox. Colors seem more vibrant. 🌈 Life more expansive. Trying to remember how RSS works again…happy that FeedBin still exists.

    → 1:34 PM, Jul 18
  • Dumping Twitter and moving back to RSS and of course, this album pops up on Spotify for me. 😄 Toad the Wet Sprocket

    → 10:04 PM, Jul 17
  • Hooking up Muse to 30 years of narrative study was really fun last week. So much so, I recorded a little introduction to what you can do with it now. The 1997 version of me would be seriously impressed. 🤗 youtu.be/ihvf2E7Kt…

    → 2:23 PM, Jul 17
  • While others are focused on unraveling the mysteries of existence 🙄, we’re over here changing what it means to develop a story!

    OpenAI added Function Calling, making it even easier to tap into Subtxt’s extensive knowledge-base. 🚀

    Revolutionizing Storytelling

    → 2:27 PM, Jul 12
  • Phenomenal use of Panning in Midjourney. It’s like getting a chance to revisit a world that simply doesn’t exist anymore—literally expanding my dreams as a child. 🤯🤗

    → 9:01 AM, Jul 9
  • This is super cool. Assisted an indie film client with Subtxt to create a treatment in 2 hrs.😅 Used GPT-4 to identify top 4-5 locations, set each of 28 scenes, and summarized events based on the treatment’s subtext. All in a day’s work. 😃

    → 5:36 PM, Jul 8
  • Discovering Intentional Storytelling

    After six months of coding with GPT-4 by my side, I can confidently say that I’ve entered a completely new world of creativity. I haven’t visited StackOverflow in months and I haven’t looked up any tutorials on YouTube. Instead, I write down how I conceptualize a function or a particular part of the app, and then work together with GPT-4 to help craft the final code.

    Earlier today, I discovered that there was a concept known as Intentional Programming. Someone had brought it up as an example of this process that I described above with the idea being that you just essentially think everything into existence.

    Intentional Programming (IP) is a concept in software development coined by Charles Simonyi from Microsoft Research. The key idea is to capture the programmer’s intent and separate it from the implementation details. Instead of writing code in a traditional programming language, the programmer works in a domain-specific language (DSL) that matches the problem space more closely.

    These “intentions” can then be transformed into a final product (usually software code in a conventional language) by the system. The advantage of this approach is that it can improve code clarity and maintenance because it emphasizes what needs to be done (intent) rather than how to do it (implementation).

    For fun, I asked GPT for to imagine a way of telling a story using something like the concept of Intentional Programming

    To my surprise, it essentially described the app I’ve been building for the last six years:

    Intentional Storytelling, as a concept, could refer to a method of narrative development in which the storyteller’s intent is separated from the actual storytelling process. This is akin to Intentional Programming, where the intent is separated from the implementation.

    In Intentional Storytelling, a storyteller would craft narratives using a Domain-Specific Language (DSL) designed for storytelling. The DSL could include elements like plot, characters, settings, themes, narrative arc, etc. The storyteller, rather than writing the story directly, would outline their intentions for these elements.

    Subtxt is like that opening scene in the first episode of the last season of Westworld, where Dolores sits down at her workstation, and essentially talks her way through her story. Subtxt works exactly the same way—just without the audio interface (yet), and without the darker undertones of what that story was all about (depending on your point of view!).

    To get even more meta, this blog post was created by something that you could call “intentional blogging”! The idea of Intentional Programming was just something I’d scrolled by on Twitter this morning. Fascinated with the idea, I went for a walk and just had a conversation with GPT-4 (through the ChatGPT app) and just kept asking questions to explore the concept. I didn’t enter text, I just talked it through using the Whisper API, (which is amazing). A couple of responses later, I wrapped it all up in a blog post that I’lll be posting up on my companies website later today.

    So there have it: Intentional Storytelling.

    → 10:18 AM, Jul 8
  • Wait. So you’re telling me the entire Star Wars Radio Drama from the 70s, the same one I obsessed over as a youngster is somehow available in its entirety on YouTube?! 🥲

    → 4:37 PM, Jun 29
  • Interesting. For all this talk about fixing “hallucinations” when dealing LLMs (there’s even an entire conference on it from Pinecone next month), I’m noticing a trend where users actually miss the more “imaginative” assistant–even if it was wrong sometimes.

    → 2:41 PM, Jun 26
  • “Save the Cat” without a Storyform is meaningless. “Save the Cat” WITH a Storyform is the opposite of that.

    → 5:12 PM, Jun 19
  • Recorded a brand new introductory video to Subtxt. I kinda like how the totally random story idea out of nowhere was starting to unfold, and will likely record follow-up episodes. Here’s the first:

    → 6:11 PM, Jun 15
  • “Steerable” is right. This is insane. Subtxt is based on some complex narrative theory which many writers find difficult to understand at times. This ONE question spawned 3-4 different concepts/ideas at once in a single response. AND encouragement for artist first! 😍

    → 9:48 AM, Jun 14
  • ok - there is a HUGE difference in the latest models. I’m telling you, if you just talk your way through a Storyform in Subtxt, you can have a meaningful narrative illustrated in NO TIME. I can’t share what I’m working on now, but I’ll see if I can find time to show how to build an entire season of television in hours. It’s crazy the difference!

    → 6:43 PM, Jun 13
  • Oh my, the improved responses with today’s API rollout from OpenAI is incredible. Even for the “lesser” powered 3.5 model, the potential for everyone to learn how to write better stories is increasing exponentially! 🤯

    → 4:51 PM, Jun 13
  • Best part about consistently blogging and writing articles on narrative structure and theory for the past 15 years? When you need to explain some concept you already wrote about in a clear and succinct manner you can just have your digital chatbot do it for you! 🤗

    → 3:14 PM, Jun 12
  • New term learned today: de-generative AI: when responses returned are meaningless and ultimately unusable. 😄

    → 11:32 AM, Jun 10
  • Absolutely loving the new iAPresenter. 😍

    → 4:49 PM, Jun 7
  • Whoever did the Snoopy animations in watchOS 10: Bravo! 👏 Really charming companion to have all day.

    → 1:50 PM, Jun 7
  • I just checked, and unless Nolan is using a pseudonym he’s not a subscriber to Subtxt. Still, this looks familiar…

    → 8:41 AM, Jun 7
  • Genuinely having a conversation with AI-chatbot I created based on my decades of researching narrative theory, and it totally feels like that opening scene with Dolores in the last season of Westworld. I wish I could share, but can’t quite yet!

    → 4:33 PM, Jun 6
  • First class of the Summer cohort finished! GREAT energy, always so much more excitement and enthusiasm this time of year. Very grateful and looking forward toward to the next six weeks. 🙏🏼

    → 1:13 PM, Jun 6
  • BIG fan of the giant email previews that pop up in Standby Mode with iOS17. Super cool!

    → 9:12 PM, Jun 5
  • This is going to be the longest eight months of my life! #WWDC23

    → 12:03 PM, Jun 5
  • This is a lot of math first thing in the morning. #WWDC23

    → 10:16 AM, Jun 5
  • There is nothing cooler than walking away from your friends in the middle of a field to start your guitar solo. youtu.be/Nx_b5nytS…

    → 4:17 PM, Jun 4
  • I generally abandoned chain-of-thought for writing stories as the results were meaningless (bad stories). But with this challenge from OpenAI…

    It is unknown how broadly these results will generalize beyond the domain of math, and we consider it important for future work to explore the impact of process supervision in other domains.

    Process Supervision might just make me a believer again.

    → 10:17 PM, May 31
  • Japan goes all in on AI, no copyright enforcement on data-training. They see it as a way to gain influence and improve their position in their world. Not sure how anyone else can afford not to do the same.

    → 1:52 PM, May 31
  • Bug Fixes and Black Holes

    App developer’s conundrum: when fixing bugs, I often just throw in some random story ideas off the top of my head as an easy way to test something. In the following, a user wrote in saying the “line-returns” in their requests weren’t being saved properly.

    I quickly opened up Muse locally, and typed in the following to test:

    I have a story about a civilization of interstellar travelers who created time through the creation of a black hole, and now, trapped in this idea of time, have to create a world to inhabit. And what they do is create Earth. Cool, right?

    What would be a good way to structure this in terms of OS Domain and what are some ideas for a MC/OC combos?

    It was based on a YouTube doc I was watching last night, and thought it would make for a cool series. I had no idea about who the Main Character would be, or the Obstacle Character (that’s the MC/OC for those not familiar with Subtxt/Dramatica).

    What it came back with was so interesting and definitely NOT what I had imagined, that now I just want to go off and write the whole thing. 😅

    (Side note: it even managed to suggest the write Domain of Universe for the story in question).

    What a great time to be alive!

    → 4:50 PM, May 30
  • AI-Driven Efficiency: The Magic of Transmuting Support Queries into Blogs for a Solopreneur

    As a solopreneur, the addition of AI to my work has unlocked massive productivity for me. Transforming support requests into documentation which are then transmutated into blog posts will never not seem like pure magic to me.

    For instance, I received this request about an hour ago about why Muse was getting parts of the Storyform “wrong.” Turns out, GPT sometimes takes narrative concepts too literally.

    I answered the question using ChatGPT, where I just conversationally typed out my thinking behind it, and why that’s happening, etc. GPT-4 took those ideas, wrapped them up in a nice email response, which I then used to reply in HelpScout.

    Knowing this would likely help others, I just asked Chat to re-write the email as a page in the documentation. Took about 30 seconds to write, 15 seconds to push to GitHub, and about a minute for Netlify to re-render the whole thing. It’s here at the bottom: Subtxt-Muse: Best Practices for Storytellers - the section labeled “When Muse Gets It Wrong.”

    As it related to a couple of posts I just did on the Discord community earlier this morning, I went ahead and copied and pasted those into Chat, and asked it to turn it into a blog post. A couple of minutes later, the whole thing was live on my company’s blog: Adaptive Storyforming: Changing Story Structure on the Fly with AI.

    Prior to AI, this process would have taken minimum three hours, and by the end I would have been too exhausted to write anything about it.

    Now, I have time to write up this quick post. 😊

    I use this technique throughout everything I do now, and can’t imagine developing an app like Subtxt without it. 🚀

    → 1:40 PM, May 30
  • Walking around my house, screaming at the top of my lungs: “I’m the eldest boy!!” whenever I want something. Works great. 😄

    → 11:07 AM, May 30
  • Blood and Gold on Netflix: super-light on story, super-fun with the spaghetti-Western nostalgia.

    → 2:06 PM, May 28
  • My twins are graduating high school today. That went fast.

    → 11:23 AM, May 26
  • Live Activities?!! What is this greatness? I’ll actually remember to stop timers now. 😄

    → 12:30 PM, May 24
  • Published a brand new article on analyzing the film “Close” with Muse, the AI digital mentor I added to Subtxt earlier this year. If you haven’t seen it yet, it’s a beautiful film (with something very important to say).

    → 11:03 AM, May 23
  • Just learned about a Klein Bottle and now I can’t stop thinking about how this relates to narrative structure. It’s only meaningful when seen through time. A 3-Dimensional Mobius strip!

    → 12:18 AM, May 23
  • The joy that resides in my heart closing all the AWS: Lambda Layers tabs from last week, never having to think about them ever again. ❤️

    → 2:54 PM, May 22
  • Wonderful meeting with a French writer this morning who is LOVING the fact that Muse automatically translates everything I’ve ever written into French, making concepts that have challenged this writer for years finally clear! 🌍

    → 12:10 PM, May 22
  • Just had a full on narrative analysis conversation with Muse, the AI story development bot I created. We were working on “Close” (beautiful film). My upgrades have drastically improved the whole experience in the last 24 hours. Feels like I’m in a university class! 🎓

    → 12:54 AM, May 21
  • Cannot tell you how rewarding it is to finally get sparse-dense embeddings working in Subtxt now. 💪😎 The difference is night and day when it comes to the terminology of narrative concepts. On another note, there are several computers now working in tandem to deliver Subtxt. 🤯

    → 11:16 PM, May 19
  • Brilliant interview w/ WA. Best part is here 👉 Interview snippet. If you were ever curious as to how Subtxt works, this is the how and why. The structure of story is the structure of our mind (the observer).

    → 2:36 PM, May 19
  • work buddy.

    → 3:41 PM, May 18
  • Every day is something new when it comes to building an AI app. Now I have customers sending me conversations with my chatbot where its specifically helping them overcome the “limitations” of the interface I created. 🤣 It legit suggested a great idea for working around what I made.

    → 11:39 AM, May 18
  • People I’ve been working with for almost a decade are finally getting the concepts because of AI. Why? AI === infinite patience in teaching and flexibility in teaching and examples. My community is quiet, but my app’s chat log is a hockey stick. 📈

    → 4:15 PM, May 17
  • Super cool! 😎 Maven makes it easy now to add your own “lead magnet” for your cohort courses. The one coming up will be my tenth in the series (!), so I made a downloadable PDF to show students what they can expect from the class. 🚀 https://maven.com/p/d1d191

    → 11:26 AM, May 17
  • A little bit of narrative theory goes a long way, especially when you have a fine-tuned super intelligence there alongside to help you out. In this latest excerpt from the Guide, I explore how to adapt a French novel using Subtxt. 🤖💪✍🏻

    → 1:39 PM, May 15
  • Happy Mother’s Day to all the wonderful mothers out there. No idea how I lucked out so much with mine, but she’s proof positive that someone somewhere is looking out for me. ❤️

    → 10:04 AM, May 14
  • Aha! I think I found what I was looking for to host some sparse embedding goodness. @streamlit looks just like what I needed to get excited about next week’s development. 🔨👷‍♂️

    → 8:14 PM, May 13
  • Growing up in the 80s, I automatically assumed HuggingFace was about “Aliens.” It’s not. It’s about the emoji. 🤗 https://huggingface.co

    → 4:35 PM, May 13
  • Watching Stand By Me, and it’s still crazy to me that this is even possible, let alone that I built it.

    → 2:45 PM, May 13
  • Shane Hawkins drumming on My Hero. 🥲youtu.be/P2KnD7sfp…

    → 12:38 AM, May 13
  • I used to draw Mickey Mouse for a living. 😆

    → 5:30 PM, May 12
  • Had a thought while running – knowing intimately now how AI works (at least, GPT type stuff), it would not be difficult at all to create an animation transformer based on the past 20 years of CG animation. All the training data is already there…🥺

    → 10:39 AM, May 12
  • Alright, time to get at that sparse/dense index for Subtxt and all that narrative theory! 💪

    → 6:44 PM, May 11
  • Can’t stop screenshot-ing Arc and everything that pops up. Beautiful! 🥰 And that opening introduction! 🥹

    → 6:35 PM, May 10
  • Access to Arc! 🎉😃

    → 6:24 PM, May 10
  • It’s so crazy how story works in our minds. A year ago, I put Booksmart into Subtxt with a certain understanding of the narrative structure. But today, thinking back, I remember the “general” (Genre) differently than what I have. All the storytelling has faded away, leaving just the structure.

    → 8:26 PM, May 9
  • AI conundrum: I love all the advances. I despise the transitional phrases (“However”, “In conclusion,”). They’re so much a part of GPT right now that I can easily see when something has been written with AI. Can’t wait till they tweak this out of existence!

    → 3:44 PM, May 9
  • I can think of no greater chasm right now than George R.R. Martin’s take on what it was like to write in the 1980s and scientists using GPT to passively read people’s thoughts. Both articles: Writers on Set & Storytelling Telepathy.

    → 12:05 PM, May 9
  • Phenomenal trailer for Oppenheimer. Why? It’s the first indication of the complete story being told here (Obstacle Character, Relationship Story, etc.). Now I’m really looking forward to it. youtu.be/uYPbbksJx…

    → 5:52 PM, May 8
  • Easter egg in the latest episode of Barry. If it’s true that the most recent is 8 years in the future, I’d say this headline is about 7 ½ years too late. 😊

    → 4:06 PM, May 8
  • It’s crazy the kind of talented writers I get to work with on a weekly basis. My only hope is that everyone gets a chance to see just how massively created these individuals are in the very near future.

    → 3:00 PM, May 8
  • Anybody worried about AI replacing writers: just watch “Aftersun”. As someone deeply entrenched in GPT and narrative structure, I can confidently say there’s no robot capable of replicating that kind of connection and honesty in storytelling. An exceptional film. 🎥

    → 10:50 AM, May 7
  • 🤯 I just discovered that the actress who plays Nebula in the Marvel movies is Amy Pond from Dr. Who! I always thought her character in Avengers was a cut above the rest—now I know why. Phenomenal actress.

    → 1:40 AM, May 6
  • “One thing that I think we will find is that biological intelligence is incredibly limited, relative to what we’re capable of producing.”

    Daily hands-on AI experience for a year: I fully endorse Altman’s comment.

    → 8:00 PM, May 5
  • What do you do when someone sends you a 1500 word support request? Easy! Ask GPT-4 to summarize the key points in a bullet list, then turn those into a QnA page for everyone to enjoy. 😊guide.subtxt.app/informati…

    → 4:33 PM, May 4
  • Leaked paper on “moats” and AI and how there’s no way what you build will remain a secret, nor will your models be a strategic advantage. Goes for those studios building out their AI🤫. It’s you and who you are that attracts your audience (i.e., your artists) not your IP.

    → 11:56 AM, May 4
  • Met with a former student yesterday who used to teach game theory at MIT and has ideas on how I could make my cohorts even better. Very grateful for all the help. 😃

    → 10:10 AM, May 3
  • I have returned to Southern California. Where are all the trees? (P.S. I miss Northern California 🌲)

    → 10:58 AM, May 2
  • First time here. 😳

    → 7:20 PM, Apr 29
  • After an hour of reading Jack’s thoughts on the Twitter sale, it’s clear to me that the skies are bluer on the other side. If anyone wants an invite, I’ve got one. 😊

    → 8:13 PM, Apr 28
  • Twitter brought me back here.

    → 11:10 AM, Apr 26
  • RSS
  • JSON Feed
  • Micro.blog