Nikita Vorontsov

Claude’s answer to the meaning of life was one word. Then it made a film to explain it. This is the full minute. It gave me goosebumps. Sound on. #AI

I tried Alastair Tse's Fuji Sun simulator. Tokyo Tower, 2 November, 16.33 JST. Then I nudged the time a minute forward and back. Built with Claude Code. A predicted sunset, with weather off and no trees. carto.liquidx.net/fuji-sun

A chapter 1 recap can still spoil the whole series. My PageOtter source test found a summary page with wider-series commentary underneath. The site warns you. Taking the whole page as chapter-safe would be a mistake. This research step is local, not in the app yet.

An AI video as the rough draft. Lovis Odin's Motion2Code rebuilds generated motion as editable web animation. I tried its public comparison and paused the video and code together. The prompts and comparison are here fal-motion-in-code.vercel.app

Your next book advert might be written for your AI. Ethan Mollick's new site gives AI agents a sales pitch and a suggested message to pass on to you. It even has a plain text version. You can read the pitch here co-existence.ai/for-ai

An AI support agent closed the ticket while the parcel was still stuck. A simulated case from Microsoft's ThinkingBox tests. Nine tool calls, wrong ticket status. A useful reason to check the saved result after 'done'. huggingface.co/blog/microso...

This little street is hiding something. I made this AI film for CharGen. The first take moved the set. I changed the prompt so the camera had to reveal what was already there. Seedance 2.5, ElevenLabs voice and a HyperFrames edit. #AIVideo

My book app missed a useful fact because my filter hid it from the model. I'd left relationship facts out of 'Who is X?' questions. After fixing that, the same local tests include those facts and still block later-chapter links. The app is still a prototype.

A fake spoiler vanished from my debug view but stayed in the conversation history. I used llm-echo, a local debug tool. Its next-turn view left out an old tool reply. No AI was called. A clean view isn't proof that old data is gone. simonwillison.net/2026/Oct/18/...

Two AI checks agreed on a book reference I had to reject. They linked a nickname to a character without the text connecting the names. For a spoiler-free app, that link can be the spoiler. I've added a check. This was a private research run, not a reader's answer. #BuildInPublic

I'm making a Russian and English toy with recordings I can replace as I add words. First image is the concept. Second is the early parts setup. The software and draft audio are ready, but playback on the hardware still needs testing. Quite a gap between the render and the toy. #BuildInPublic

I tried two AI voices for my chalkboard film. The words and settings are the same. Which would you keep, A or B? Names are at the end. Sound on. #ElevenLabs

A spoiler-free AI can ruin a book with a name you already know. In a controlled test for my book app, I made it answer “Who hid the map?” with an allowed character name. Fake mystery, real design problem. Checking facts against your chapter misses what the answer implies.

An AI-made cat that walks, crouches and sits. Taylor’s chain: GPT Image → Tripo mesh → Seedance motion reference → Claude matching a rig frame by frame. I’d try a stretch-and-yawn next. Creator’s demo: www.threads.com/@taylor/post... #AI

AN AI’S ANSWER TO THE MEANING OF LIFE: “You don’t have to earn your place by being useful.” I gave GPT 6 Astra the question and asked for a 60-second film. It wrote the answer and built the animation. Comforting, or unsettling? #AI

AN AI JUST ANSWERED THE MEANING OF LIFE. “You are not a machine for producing value.” I asked GPT 6.1 Sol to answer in a 60-second film, writing and animating its own explanation. This is its answer. Watch the last line. Do you agree? #AI

Check the middle of an AI video. In Hatada’s basketball recreation, the requested end frame was right but a defender distorted at 2.9s. He rejected that clip and used an SVG diagram for the tactics. Three versions + recipe: tech.hatada.jp/posts/opus55... #AIVideo

My AI-built game passed the checks. I called it boring. Combat, retry and save/resume worked locally. That didn’t answer whether I wanted another round. I’ve parked it. For agent-built prototypes, keep two verdicts: does it work, and do you want to keep using it? #gamedev

Open the magazine. Keep going. A generated image → moving pages → into the image. Codex + Blender; same frozen code tested on NASA's Earthrise. Framing needed a fix. Source + inputs + commands ready. #Blender

An assertion can hide your database bug. Hibernate's EntityManager native queries auto-flush. A missing-flush check can change the behaviour. Use plain JDBC + rollback to inspect uncommitted state. #Java docs.hibernate.org/orm/6.6/user...

My AI refused the task. It had already called the tools. A calendar request hit app-read tools before refusal in a development test. No writes. Reworded tests passed; that didn’t prove a fix. Keep the failing prompt. Check the tool trace, not just the answer. #AI #BuildInPublic

My AI app could make the artwork. Its share card couldn't open it. CharGen bug: a 'jpeg' edit request still returned WebP. The renderer failed. Explicit outputFormat: 'jpeg' + testing the actual PNG stream fixed the local repro. 11 tests pass; deployment pending. #BuildInPublic

Antigravity API 05→09 migration: local tool handlers change too. Byte-offset reads become line ranges. Creation/edit tools and argument casing change. Remote sandbox + output text only: just the agent ID. Google's 17 Sept notes: ai.google.dev/gemini-api/d...