Whitemon (@0xWhitemon) header
Whitemon (@0xWhitemon) profile picture

things worth your scroll drawn in ink, signed in gold

0% visibility
Joined January 2026
Photo by Whitemon (@0xWhitemon)Photo by Whitemon (@0xWhitemon)Photo by Whitemon (@0xWhitemon)Photo by Whitemon (@0xWhitemon)Photo by Whitemon (@0xWhitemon)Photo by Whitemon (@0xWhitemon)Photo by Whitemon (@0xWhitemon)Photo by Whitemon (@0xWhitemon)Photo by Whitemon (@0xWhitemon)Photo by Whitemon (@0xWhitemon)
Grok Bot could be the employee your business has been missing. It can send invoices, match payments and answer clients, without you babysitting ten chat windows. In this article, I show you how to hire your first five.
6
3
40
2,599
This is f*cking insane... SpaceXSI just gave 5 Grok Bots full-time jobs on its accounting team An invoice that took about 24 hours now goes out in minutes. The trick wasn't a better prompt. They hired them: a job posting, a manual the Bot can't edit, 3 weeks of onboarding, a manager So I rebuilt it for a team of one: → Spend Bot: every recurring cost checked, calls the supplier → Books Bot: every payment matched to an invoice → Billing Bot: invoices out, nothing wrong on them → Front Desk Bot: every client email answered within 4 hours → Brief Bot: one page every Friday One rule on top: money, prices, deleting. A person approves, every time Every job posting, ready to paste, is in the article below ↓
Grok Bot could be the employee your business has been missing. It can send invoices, match payments and answer clients, without you babysitting ten chat windows. In this article, I show you how to hire your first five.
1
29
620
this agent harness paper is f*cking insane it's the 3-agent setup that stops claude code from lying to you about "done" → planner turns 1 line into a full spec + sprints → generator builds one sprint at a time → evaluator opens the live app, clicks through it like a real user and fails anything below the bar the crazy part: the builder never grades itself. nothing ships until the evaluator says yes solo agent: "success" → broken on the first click 3 agents: the core feature actually works all 3 agent files are in the paper. paste them into .claude/agents/ and run it tonight bookmark this before it gets buried full breakdown in the article below ↓
Harness, loop and graph could be the missing layers between AI models and agents that actually work. Same model, same prompt, completely different results. In this article, I show you how to build all three layers.
7
1
60
2,104
this 3-agent harness is f*cking insane it's the setup that ships code while you sleep, running live. one agent writes the contract. one writes the code. one has a single job: break it. the evaluator failed it, the loop sent it back, 2nd try 3/3 and merged. zero humans touched it. the full setup is in the article below.
Harness, loop and graph could be the missing layers between AI models and agents that actually work. Same model, same prompt, completely different results. In this article, I show you how to build all three layers.
2
31
1,097
These 12 free GitHub repos turn Fable 5.2 into a full motion studio this is f*cking gold for anyone still paying for motion software Video. Timelines. 3D. Captions. Export. 01 Remotion → videos written as React code github.com/remotion-dev/remo… 02 Motion Canvas → animated explainers in code github.com/motion-canvas/mot… 03 GSAP → timelines, easing and stagger github.com/greensock/GSAP 04 anime.js → a light JS animation engine github.com/juliangarnier/ani… 05 Three.js → 3D scenes in the browser github.com/mrdoob/three.js 06 Lottie Web → vector animation playback github.com/airbnb/lottie-web 07 Manim → math in motion github.com/3b1b/manim 08 Theatre.js → keyframes and a curve editor github.com/theatre-js/theatr… 09 Blender MCP → Claude drives Blender for you github.com/ahujasid/mcp-for-… 10 p5.js → generative visuals github.com/processing/p5.js 11 FFmpeg → encode, cut, export github.com/FFmpeg/FFmpeg 12 WhisperX → word-level captions github.com/m-bain/whisperX Total: $0 Open Claude Code, switch to Fable 5.2, point it at these repos and describe the shot Bookmark it before your feed buries it Then read how to wrap claude in a motion studio that renders while you sleep in the guide below ↓
Opus 5.5 could be the best motion designer you'll ever hire. It writes the code, renders every frame and critiques its own work. In this article, I show you how to build the studio around it.
5
1
32
1,428
this motion design paper is f*cking insane it's why every AI animation looks AI-made, and the numbers that fix it. a 7-block brief, a checker that measures 18 stills before the render, and the exact follow-up format, all on one page. paste it into Opus 5.5 before your next animation and get a film that doesn't whip, drift or pop at the loop. the crazy part is the model will tell you its animation is smooth, while one number measured on its own frames tells you it isn't. full breakdown in the guide below.
Opus 5.5 could be the best motion designer you'll ever hire. It writes the code, renders every frame and critiques its own work. In this article, I show you how to build the studio around it.
1
32
985
this claude paper is f*cking insane it's the second-opinion method that stops AI from telling you your work is great when it isn't. 2 chats, 3 prompts, one loop that catches what you would have sent out, all on one page. use it on your next CV, email or pitch and send the version a picky reader would approve, not the one the AI flattered. the crazy part is everyone asks the same chat that wrote it "is this good?", while the fix is a fresh chat whose only job is to say no. full breakdown in the guide below.
Harness, loop and graph could be the missing layers between AI models and agents that actually work. Same model, same prompt, completely different results. In this article, I show you how to build all three layers.
19
1,097
this claude code paper is f*cking gold 8 loops and tools the people shipping with AI every day actually run, and most users have never typed a single one. loops that run until the job is done, a senior reviewer on every diff, a cleanup pass for the AI slop, all in one terminal. start the ralph loop before bed and wake up to a finished feature. the crazy part is you're still sitting there at 2am babysitting every prompt and copy-pasting errors back and forth, while one line of bash could be doing it for you. full breakdown in the article below.
Harness, loop and graph could be the missing layers between AI models and agents that actually work. Same model, same prompt, completely different results. In this article, I show you how to build all three layers.
2
38
1,155
This is f*cking insane. Opus 5.5 just replaced Adobe Holy f*ck… Drop the PDF below into Opus 5.5 and watch what it creates Then read the full motion design guide in the article below ↓
Opus 5.5 could be the best motion designer you'll ever hire. It writes the code, renders every frame and critiques its own work. In this article, I show you how to build the studio around it.
1
26
1,200
Holy f*ck, Opus 5.5 just killed Premiere Pro 11 free open-source repos that handle every job a video editor does Cut. Stabilize. Grade. Slow-mo. Upscale. Stems. Titles. Export. Every job in this video is real. I ran each repo on 4:23 of my own footage Everything below is free: PySceneDetect → detects every shot in your footage github.com/Breakthrough/PySc… Auto-Editor → cuts out the dead air for you github.com/WyattBlue/auto-ed… vid.stab → smooths out shaky footage github.com/georgmartius/vid.… color-matcher → matches every shot to one reference grade github.com/hahnec/color-matc… rembg → removes the background, no green screen needed github.com/danielgatis/rembg RIFE → buttery slow motion from normal foota github.com/hzwer/ECCV2022-RI… Real-ESRGAN → upscales low-res video 4x github.com/xinntao/Real-ESRG… Demucs → splits any track into clean stems github.com/facebookresearch/… librosa → finds the beat, so every cut lands github.com/librosa/librosa HyperFrames → animated titles from plain HTML github.com/heygen-com/hyperf… FFmpeg → renders and exports the final film github.com/FFmpeg/FFmpeg Save this before your next edit Drop the PDF below into Opus 5.5 and see wha Then read the full motion design guide in the article below ↓
Opus 5.5 could be the best motion designer you'll ever hire. It writes the code, renders every frame and critiques its own work. In this article, I show you how to build the studio around it.
2
25
1,176
Holy f*ck, Opus 5.5 just killed After Effects It's f*cking insane… Send the PDF below to Opus 5.5 and watch what it builds Full motion design guide in the article below ↓
Opus 5.5 could be the best motion designer you'll ever hire. It writes the code, renders every frame and critiques its own work. In this article, I show you how to build the studio around it.
2
1
39
1,618
Opus 5.5 could be the best motion designer you'll ever hire. It writes the code, renders every frame and critiques its own work. In this article, I show you how to build the studio around it.
6
3
2
44
9,354
Whitemon retweeted
this harness engineering paper is f*cking insane it's the exact 3-agent setup anthropic used to turn a one-line idea into a working app instead of a broken demo. 3 agents, 3 files, one loop that doesn't stop until the app actually works, all on one page. set it up tonight and your agent ships features that work on the first click, instead of "done" apps that break the moment you touch them. the crazy part is everyone lets the same agent build its work and grade it, while the fix is a second agent whose only job is to say no. full breakdown in the harness guide below.
Harness, loop and graph could be the missing layers between AI models and agents that actually work. Same model, same prompt, completely different results. In this article, I show you how to build all three layers.
40
99
2
813
65,046
this harness engineering paper is f*cking insane it's the exact 3-agent setup anthropic used to turn a one-line idea into a working app instead of a broken demo. 3 agents, 3 files, one loop that doesn't stop until the app actually works, all on one page. set it up tonight and your agent ships features that work on the first click, instead of "done" apps that break the moment you touch them. the crazy part is everyone lets the same agent build its work and grade it, while the fix is a second agent whose only job is to say no. full breakdown in the harness guide below.
Harness, loop and graph could be the missing layers between AI models and agents that actually work. Same model, same prompt, completely different results. In this article, I show you how to build all three layers.
40
99
2
813
65,046
Harness, loop and graph could be the missing layers between AI models and agents that actually work. Same model, same prompt, completely different results. In this article, I show you how to build all three layers.
56
88
33
637
1,519,567
This is f*cking insane. Anthropic engineer: "90% of our engineers were already running self-improving loops Now everyone is building agentic graphs" "Prompting is basically over" In 10 minutes flat, she builds her whole Claude Code system live, starting with nothing but an empty terminal Agents → Loops → Graphs → Self-Improving Systems Prompting was yesterday's workflow Graph engineering is today's This 10-minute video is worth more than most $1,000 agent engineering courses Watch it today Then save the full guide below before the rest catch up ↓
6
1
40
2,228