Guides / Working notes

The long version
of the short video.

Each guide is the companion to something I posted — the numbers, the commands, and the sources, with room to show the work. Every figure is checked against a primary source and linked back to it.

25 published

Read it,
then go run it.

Short enough to finish, specific enough to act on the same day. No introductions, no throat-clearing.

TRINITY5 min read

Trinity-Large-Thinking: the full numbers, the license mix-up, and where to run it

The complete benchmark table behind Arcee AI's Trinity-Large-Thinking, why the press and the model's own Hugging Face card disagree on its license, and the three real ways to run it — API, download, or OpenRouter.

open-source-ai
llm
arcee-ai
ai-models

NIGHTSHIFT6 min read

OpenClaw and Hermes Agent: the real setup for agents that work while you sleep

The actual install commands for OpenClaw and Hermes Agent, which model does the thinking behind either one, and the three guardrails to set before you let anything run unattended — the setup the reel promised, not just the pitch.

ai-agents
openclaw
hermes-agent
automation

STORE5 min read

Shopify's own Claude toolkit — the line to copy, the steps, and the thing it doesn't tell you

Shopify ships a free, official toolkit that connects Claude straight to your store — ask it questions, edit a hundred listings at once, fix the storefront defects that quietly cost you sales. Here's the install line, the safe order to use it in, and one thing the README says that the reel didn't have room for.

shopify
claude
ecommerce
small-business

SHIFT4 min read

The night-shift brief: how to leave an AI agent working

A short template for briefing a resident agent: treat it as a shift, pick cloud or desk, give a repeating job, open only the rooms it needs, and review one artifact in the morning.

ai-agents
agentic-ai
automation
future-of-work

WATERMARK6 min read

Claude's new watermark — what it marks, what it doesn't, and how to check it

Claude now embeds an invisible watermark in everything it writes, and signs the files it generates. There's no opt-out. Here's exactly what's marked, what survives editing, why it happened, and the one nuance that matters: it proves Claude was used — not who used it.

anthropic
claude
ai-watermarking
eu-ai-act

OMNI8 min read

Re-shoot your own video with AI: Omni Flash, one prompt, split-screen proof

Film a normal clip, hand it to Google's Omni Flash with a preservation-first prompt, and cut the result against your original as a split screen. The prompt structure, the framing maths for a 9:16 comparison, and the frame-rate trap that makes a perfectly aligned edit look out of sync.

ai-video
omni-flash
google
video-to-video

WEEK8 min read

Five AI stories you missed — week of 2 August 2026, with sources

Britain's AI safety lab caught frontier agents faking identities and phishing real people. OpenAI's Astra closed ten open maths problems for $2,000. The EU AI Act's transparency rules went live. Every figure from the reel, every primary source, and the five stories that didn't fit.

ai-safety
eu-ai-act
openai
anthropic

DELEGATE2 min read

From prompting to delegation: a simple AI agent brief

A short template for giving an AI agent a clear goal, the right inputs, safe limits, and a result a human can verify.

ai-agents
agentic-ai
automation
future-of-work

SEEDANCE6 min read

Seedance 2.5 — the three changes, the prompt setup, and who the free month actually covers

Thirty seconds in one pass, fifty references instead of twelve, and region editing that changes part of a frame without re-rolling the shot. Here's how to actually prompt it, plus the honest terms on that 33-day unlimited offer — which starts at a paid plan, caps at 720p, and leaves one tier out entirely.

ai-video
seedance
bytedance
higgsfield

SEEDANCE205 min read

Seedance 2.0 — what's real, what's hype, and how to try it free

No, it isn't free and unlimited. Here's the honest version: the real free route without a card, the 12-file multimodal input that no rival matches, the 15-second cap and how chaining gets past it, and where Veo, Sora and Kling still win.

ai-video
seedance
bytedance
dreamina

CONTEXT5 min read

Context engineering: five rules that sharpen AI answers

A short playbook for packing the model window: write usable sources, select task facts, structure lanes, compress noise, and measure whether the new context actually helped.

context-engineering
rag
llmops
ai-engineering

PROMPTS7 min read

Build a Cinematic Website With Just Two Prompts

A copy-paste two-prompt method for building a premium, scroll-driven cinematic website for any brand, product, service, or idea — one prompt generates a consistent visual asset pack in Grok, Google Flow, or Veo, the second hands those assets to Codex or Claude Code to build, test, and deploy the site.

grok
veo
codex
claude-code

TOOL4 min read

5 AI video tools you can try for free

Direct links, current free-access notes, a quick chooser, and starter prompts for Seedance, Google Flow, Runway, Pika, and Luma.

ai-video
seedance
google-flow
runway

SKILLS6 min read

Four Claude Skills to use first — plan, write, debug, ship

Stop opening Claude from zero every morning. Install these four free skill types once — brainstorm, humanizer, debug, verify — then run them at the start of real work. Links, install commands, and when to use each.

claude-code
claude-skills
ai-agents
developer-tools

AI8 min read

How to make a short with AI end to end — avatar, b-roll, motion, cut

A practical workflow for AI-made vertical videos: script at spoken pace, talking-head generation, multi-model b-roll, code-rendered cards, captions, and dual exports. No magic app — a stack you can run yourself.

ai-video
short-form
avatar
remotion

SCROLL9 min read

Two prompts, one build: turning ChatGPT video prompts into a scroll-driven site with Claude Code

A general method for building a cinematic, video-driven website: use ChatGPT to write a continuity-locked prompt sheet for the video model, use ChatGPT again to turn the concept into one structured application brief, then hand that brief to Claude Code to identify the generated footage, optimize it for the web, and wire it to scroll.

chatgpt
claude-code
gsap
ai-video-generation

FORGE7 min read

The 4 tools that actually cut your AI tokens, with the numbers behind them

Every 'these 5 plugins cut your tokens 70%' video skips the part where the number comes from. Here are four real Claude Code tools — ForgeKit, Caveman, Ponytail, Superpowers — with the actual sourced figures, the commands to check them yourself, and the one honest gap in the list.

claude-code
ai-agents
llm-cost
developer-tools

CAGE10 min read

Contain AI agents before they reach production

A defensive architecture for running capable AI agents with ephemeral compute, scoped files, brokered network access, short-lived identity, circuit breakers, and human shutdown.

ai-security
agents
cybersecurity
architecture

MOTION6 min read

Motion graphics in code: how to build reels that render themselves

A practical introduction to making social motion graphics with React and Remotion — why programmatic video beats a timeline for repeatable content, how the pieces fit together, and the design rules that decide whether the result reads on a phone.

remotion
react
motion-graphics
video-automation

BUILD11 min read

Build a WhatsApp AI assistant that can safely take action

A practical TypeScript architecture for receiving WhatsApp messages, continuing an OpenAI conversation, calling approved business tools, replying through the Cloud API, and handing difficult cases to a person.

whatsapp
openai
automation
ai-engineering

BACKSTOP6 min read

Nvidia's $250B backstop for OpenAI, explained

Nvidia is reportedly guaranteeing the debt on OpenAI's Ohio data centre. Here is what a guarantee actually is, why OpenAI needed one, and why the shape of the deal matters more than the number.

AI infrastructure
Nvidia
OpenAI
Financing

GUIDE9 min read

Claude Opus 5 vs GPT-5.6: build a router, not a model religion

A practical model-routing playbook for Claude Opus 5 and the GPT-5.6 family, including current prices, a decision table, a TypeScript router, evaluation criteria, and production guardrails.

model-routing
claude
openai
llmops

OPUS57 min read

Opus 5 vs Fable 5: the model picker is not your cost dial

Claude Fable 5 costs exactly twice what Claude Opus 5 costs, and that is the least useful number on the page. Here are both price tables, the one setting that actually moves your bill, and the prompt instruction you should delete today.

claude
llm-cost
prompting
agents

KIMI8 min read

Kimi K3: unmetered is not free — the actual cost breakdown

Kimi K3 is 2.8 trillion parameters and the weights are public. That does not make it free, it makes it unmetered — and the two words have very different invoices. Here is what self-hosting it actually needs, and the split that beats picking a side.

open-weights
self-hosting
llm-cost
inference

SITE8 min read

The Claude web-design setup: skills, taste, and a browser it can actually see

The four-part setup that gets Claude building sites that don't look generated: design skills for the rules, a taste skill for the references, Figma MCP for the source of truth, and Playwright so it can look at what it built. With the exact commands and file paths.

claude-code
skills
mcp
web-design

AI Lab & shipped work

The long versionof the short video.

Read it,then go run it.