All desks

Codex's hidden superpower: threads that spawn threads, talk to each other, and orchestrate themselves. Hamel demos running 16 parallel coding tasks with computer use and mobile control.

A practical framework for picking an OCR model: text-blocks vs document structure, API vs self-host, and why you should test on your own data instead of trusting benchmarks.

The anti-generic Claude Design workflow: prepare three files (design.md → design system → template) so every deck starts on-brand and every feedback compounds into future designs.

Full Claude Design platform walkthrough: expanded templates, bundled usage, design systems for on-brand output, and a motion-graphics-for-videos use case that is genuinely surprising.

Lucas Rocha, an engineer at Brazilian edtech Nova Escola and alum of Hamel Husain's AI Evals course, tells his evals rollout story backwards: from a team that now runs LLM judges with no tickets and no engineer in the loop, back to the messy first spreadsheet and the three barriers — data, methodology, and judgment — that every team hits along the way.

Wayfinder is Matt Pocock's pre-spec orchestration tool: it charts a map of parallel decision/research/prototype tickets, spawns agents, and resolves open questions into a spec — demonstrated live building a TikTok creator feature.

Abi Aryan breaks down where LLM latency actually goes: the compute-bound prefill phase vs the memory-bandwidth-bound decode phase, and why decode — not prefill — dominates latency and cost, especially in multi-step agentic systems. A live Colab walkthrough with TinyLlama on a T4 measures both phases across chat, RAG, creative, and agentic request shapes.

Before worrying about evals, ask if you are building the right thing: AI that dumps an unverifiable answer creates more work. Hamel walks through design patterns — provenance, progressive disclosure, expert signals, modular steps — across three worked examples.

A walkthrough of Codex desktop app remote control: connect your agentic-coding servers, kick off sessions from your phone, and watch agents work anywhere — with live session views, subagent summaries, and follow-up prompts.

Part 1 of Hamel and Shreya's AI product engineering series: the analyze → measure → improve eval lifecycle, and three mistakes to avoid when using AI to automate evals.

Hamel argues that hiring a consultant to implement AI and walk away is almost always a mistake — the underlying problems remain and the AI itself is designed to remove the middleman. Consultants work only for education or when they have skin in the game.

When a product seems impossible to eval, that is usually a sign the product itself is weak. Hamel suggests infusing user-facing signals into the product so users — and you — can assess quality.

Hamel Husain counters the 'evals take too long' objection with one principle: if you can't eval your product, your users can't either — so evals are how you figure out how to convey trust in AI answers.

Hamel Husain explains how to scientifically measure an LLM judge's noise against ground-truth human labels — and why every LLM judge is a black-box classifier with edge cases, so you have to decide how much noise is tolerable.

Hamel Husain argues that ROUGE/BLEU string-similarity scores and 1-to-10 rating scales are holdovers from traditional NLP that make evals harder to act on — nobody knows what a 4.2 means, and binary pass/fail evals are what actually let you ship.

Hamel Husain explains why he converts every 1-to-5 rating scale eval into binary pass/fail checks, scoping each failure into a specific, actionable criterion instead of hiding noise behind middle values.

Sean Kochel demos five relatively new GitHub tools for AI-assisted coding: an architecture-diagramming draw.io skill, an anti-over-engineering auditor, a free speech-to-text dictation app, shadcn's codebase audit skill, and NVIDIA's skill security scanner.

Hamel Husain on evals for non-experts: why generic metric dashboards actively hurt you, how error analysis beats top-down guesswork, what guardrails and MCPs actually are, and the argument that AI competency can never be outsourced.

Paul Ford (former Harper's editor, co-founder of Abort) and Dan Shipper process the week Opus 4.5 landed in Claude Code — the step change, the grief, the discourse, and what actually matters.

Horace He's deep dive on why LLM inference isn't reproducible even at temperature 0: the real culprit isn't GPU concurrency but batch-size-dependent kernels — server load changes your batch, batch changes your numerics. Plus batch-invariant kernels (RMSNorm, matmul, attention) that make 1000 identical requests produce 1000 identical completions.

WHO connects gambling losses to essential household spending, unequal exposure and harms to people who never place a bet.

The field's definitive reframe: gambling harms are commercial and social determinants of health, not individual failings. Global consumer losses are projected to approach $700bn by 2028, and roughly 80 million adults already live with gambling disorder.

Household transaction data across 230,171 US households and 26 state legalizations: online sports betting cut net investment deposits by about 14%, and the damage concentrated in households already living below median savings.

England's evidence review finds the lowest participation but the highest levels of harmful gambling among the most disadvantaged groups.

Jim Bell's 1996 'Assassination Politics' essay designed a market where you bet on when someone will die and collect by doing the killing. Brian Merchant reported on the dark-web sites that tried to build it — and on the people who died while law enforcement looked away.

Lottery spending displaces other household consumption, with larger proportional reductions among low-income households.