25 posts
Boris Cherny hasn't written a line of code in 2026. He writes loops that write code. This is a field guide to the species of the prompt safari, the loops that replaced prompting, and why building the machine matters more than feeding it.
Factory just said the quiet part out loud. Engineers don't write software anymore. They build the factories that write the software. Here is what that actually means.
A single command that decomposes a plan, spins up parallel sub-agents in isolated git worktrees, runs QA, tracks costs, and delivers a tabbed HTML report. Here is how /buildthis works, why I built it, and how to install it.
Boris Cherny published a framework for AI adoption that maps the journey from gated access to thousands of autonomous agents. Most engineering teams think they are further along than they are. Here is how to tell where you actually stand.
For most of computing history, thinking was free. In 2026 it got a price tag. We metered cognition, and every metered utility eventually gets gamed.
Meta ranked 85,000 staff by token use. Amazon and Uber followed, then all scrapped the leaderboards. The best engineers code with intent, not volume.
Lightrun surveyed 200 SRE and DevOps leaders: 43% of AI code needs manual debugging, 0% are very confident, and Amazon lost 6.3M orders to it.
Opus 4.7 took #1 on Arena, SWE-bench jumped 7, vision tripled. But MRCR fell 46 points and the tokenizer quietly inflated costs. What benchmarks miss.
Eigen Labs turns idle Macs into AI compute nodes where the owner cannot see your prompts. How the private-inference network works, and what breaks.
GPT-5.4-Cyber finds bugs and chains exploits. OpenAI gives it to thousands of vetted defenders; Anthropic locked its version to 40 orgs. Two bets.
A new IETF draft proposes IPv8: 64-bit addresses, IPv4 as a subset, 8 services in one Zone Server. What is genuinely novel, and what is fantasy.
World Labs open-sourced Spark 2.0: a streaming Gaussian Splatting engine rendering 100M+ splat worlds in any browser via Three.js and WebGL2.
Cortical Labs grew 200K neurons on a chip, taught them to play Doom, then wired them into an LLM. The neurons now override token selection in real time.
DORA surveyed 5,000 devs. AI makes good teams great and broken teams worse. Here is the 7-capability playbook that separates the two.
PwC surveyed 1,217 execs across 25 sectors. The top 20% take 74% of AI value, and the gap is not budget. It is strategy. Here is what they do.
One system prompt that turns any LLM into a senior production-readiness reviewer: 17 categories, PASS/YELLOW/RED, evidence-backed. The gate before you ship
A Claude Code skill that makes AI talk caveman cut output tokens 65%, 87% on React debugging. Why brevity makes LLMs cheaper and sharper.
A new WICG proposal renders real, interactive HTML into canvas. The three primitives, privacy-preserving painting, and WebGL support, explained.
Microsoft open-sourced 7 MIT-licensed packages for AI agent governance: identity, policy enforcement, trust scoring, and full OWASP coverage at sub-0.1ms.
Meta paid $14.3B for 49% of Scale AI, built Muse Spark from scratch in 9 months, and went closed-source. Every benchmark, trade-off, and what it means.
Jensen Huang made it clear at GTC 2026: the future of AI is not massive proprietary models. It is small, open, and everywhere.
The models already have the raw capability. The jagged intelligence frontier is not about the AI — it is about you.
Jensen Huang's GTC quote decoded: why AI spending is the new engineering budget baseline.
10 problems. 45 seconds. Got #1 via an exploit discovery. Here is the full breakdown.
Loading a skills.md into context is just prompt engineering. It does not change the model. Real skills need a different architecture.