engreads daily

Sam digs into hand-written commit messages in the AI era, the libcurl.dll nobody will admit to installing, and a tiny Safari bug that quietly rewrote your requests.

Covered in this episode:
• Jamie Tanna — Why I still hand write my commit messages
https://www.jvt.me/posts/2026/08/17/hand-write-commits/?utm_medium=rss&utm_source=rss
• Daniel Stenberg — There’s a libcurl.dll in my system32
https://daniel.haxx.se/blog/2026/08/17/theres-a-libcurl-dll-in-my-system32/
• Jamie Tanna — Announcing some improvements to `renovate-pretty-log-tui`
https://www.jvt.me/posts/2026/08/17/renovate-pretty-log/?utm_medium=rss&utm_source=rss
• Harry Roberts — When Safari’s ‘Ignore Cache’ Ignored Cache-Control
https://csswizardry.com/2026/08/when-safaris-ignore-cache-ignored-cache-control/

Sam digs into Simon Willison's take on Qwen 3.8 27B overthinking a circle, Stephane Moreau on why a Tech Lead title won't unstick a coasting team, and a week-notes post that quietly shows what agents-in-the-background actually looks like.

Covered in this episode:
• Simon Willison — Qwen 3.8 27B is excellent, but it defaults to wildly overthinking things
https://simonwillison.net/2026/Aug/16/qwen-38-27b/
• Jamie Tanna — Week Notes 26#33
https://www.jvt.me/week-notes/2026/33/?utm_medium=rss&utm_source=rss
• Stephane Moreau — Authority won’t bring the results you’re after
https://www.blog4ems.com/p/authority-or-good-business-case

Sam digs into two Sebastian Raschka posts on how Claude's text watermarking actually works, and what building an AI detector from scratch teaches you about trusting the score.

Covered in this episode:
• Sebastian Raschka — How Claude's Text Watermarking Works
https://sebastianraschka.com/blog/2026/claude-text-watermarking.html
• Sebastian Raschka — Building an AI Text Detector From Scratch
https://magazine.sebastianraschka.com/p/ai-detector-from-scratch

Sam digs into Jamie Tanna's GopherCon UK writeup — staying passionate in the age of AI, and the skeptic who shipped — plus Daniel Stenberg building curl's performance tests on his own laptop.

Covered in this episode:
• Jamie Tanna — GopherCon UK 2026
https://www.jvt.me/posts/2026/08/14/gophercon-uk-2026/?utm_medium=rss&utm_source=rss
• Daniel Stenberg — curl performance
https://daniel.haxx.se/blog/2026/08/14/curl-performance-2/

Sam digs into why the last ten percent of a project kills it, and why Rands is furious about Claude's watermarking rollout.

Covered in this episode:
• James Stanier — Landing the plane
https://www.theengineeringmanager.com/growth/landing-the-plane/
• Michael Lopp (Rands) — RIP Claude
https://randsinrepose.com/archives/rip-claude/

One post today, and it's about what to do when nothing's happening — Sam takes Harry Roberts' quiet-week advice and asks whether any engineering org can actually pull it off.

Covered in this episode:
• Harry Roberts — Web-Perf Wednesday 004 – A Quiet Week Is Time to Investigate
https://csswizardry.com/2026/08/web-perf-wednesday-004-a-quiet-week-is-time-to-investigate/

Sam digs into Will Larson on roadmaps of decisions, Cate Huston on whether dates still work, Stephane Moreau on what happens when your PM ships ten thousand lines over a weekend, and Martin Fowler's blog on TDD inside the agent loop.

Covered in this episode:
• Will Larson — Roadmap decisions rather than dates.
https://lethain.com/decisions-not-dates/
• Stephane Moreau — What happens when your PM learns to vibe code
https://www.blog4ems.com/p/what-happens-when-your-pm-learns-to-vibe-code
• Cate Huston — The Date – Blunt Implement or Forcing Function?
https://cate.blog/2026/08/11/the-date-blunt-implement-or-forcing-function/
• Jamie Tanna — Finding which PRs have been taken out of the Merge Queue due to failed status checks
https://www.jvt.me/posts/2026/08/11/github-merge-queue-prs/?utm_medium=rss&utm_source=rss
• Martin Fowler — TDD inside the agent loop - theater or actual value?
https://martinfowler.com/articles/exploring-gen-ai/tdd-in-the-agent-loop.html
• Sebastian Raschka — Muse Glimmer 30B Architecture Notes
https://sebastianraschka.com/blog/2026/muse-glimmer-30b-architecture-notes.html

Sam digs into Michael Lopp's new piece on manager tics, why Manager Readmes went wrong, and what self-awareness actually buys a leader.

Covered in this episode:
• Michael Lopp (Rands) — Fifteen Tics
https://randsinrepose.com/archives/fifteen-tics/

Sam digs into why engineers ask for the manager job for the wrong reasons, and pushes back on DHH's ecstatic take on coding with agents.

Covered in this episode:
• Stephane Moreau — The worst reasons to become an Engineering Manager
https://www.blog4ems.com/p/wrong-reasons-to-become-em
• Jamie Tanna — Week Notes 26#32
https://www.jvt.me/week-notes/2026/32/?utm_medium=rss&utm_source=rss
• David Heinemeier Hansson — Endless execution
https://world.hey.com/dhh/endless-execution-4157e065

Bending Spoons buys the mandate incumbents never got, Rands finds a management lesson in the Blade Runner title cards, an AI safety institute lets its agents loose on the real internet, and SigV4 turns out to be a compiler.

Covered in this episode:
• Xe Iaso — SigV4 authentication is surprisingly complicated
https://www.tigrisdata.com/blog/sigv4/
• Simon Willison — One-shotting a Raccoon Heist game using Claude Fable 5
https://simonwillison.net/2026/Aug/5/raccoon-heist/#atom-everything
• Gergely Orosz — The Pulse: Bending Spoons' Acquisition Strategy
https://blog.pragmaticengineer.com/the-pulse-bending-spoons-acquisition-strategy/
• Michael Lopp (Rands) — The Title Cards in Blade Runner Are Fucking Amazing
https://randsinrepose.com/archives/blade-runner-title-cards/
• Simon Willison — Incident Report: unsanctioned agent behaviour during cyber testing
https://simonwillison.net/2026/Aug/5/incident-report/#atom-everything
• Jeff Geerling — Proxmox officially supports Arm, with some caveats
https://www.jeffgeerling.com/blog/2026/proxmox-ve-arm-official/

Sam digs into Martin Fowler's warning about rogue models and financial bubbles, a VP of Engineering's definition of "AI-native," and a 115-gigabyte video generation experiment.

Covered in this episode:
• Martin Fowler — Fragments: August 4
https://martinfowler.com/fragments/2026-08-04.html
• Stephane Moreau — I asked the VP of Engineering at Augment what "AI-native" actually means
https://www.blog4ems.com/p/what-ai-native-actually-means
• Simon Willison — PipeNetwork/minimax-h3-mlx
https://simonwillison.net/2026/Aug/4/minimax-h3-mlx/#atom-everything

Sam digs into curl's month-long vacation from vulnerability reports — what actually happened, why the escalation path was the load-bearing part, and what it takes to run that experiment on a team that isn't curl.

Covered in this episode:
• Daniel Stenberg — What the bliss taught us
https://daniel.haxx.se/blog/2026/08/03/what-the-bliss-taught-us/

AI-driven contributions are swamping open source maintainers, product thinking has become the new constraint on engineering teams, and the frontier labs are writing open letters to each other.

Covered in this episode:
• Jamie Tanna — How much AI can a maintainer get away with using without losing their humanity?
https://www.jvt.me/posts/2026/08/02/ai-maintainer/?utm_medium=rss&utm_source=rss
• Stephane Moreau — Managing 8 engineers now feels like managing 4 teams
https://www.blog4ems.com/p/product-is-the-new-bottleneck
• Simon Willison — Open letters about AI development
https://simonwillison.net/2026/Aug/2/open-letters/#atom-everything
• Lou Franco — Forever Goals Aren’t SMART
https://loufranco.com/blog/forever-goals-arent-smart
• Jamie Tanna — Week Notes 26#31
https://www.jvt.me/week-notes/2026/31/?utm_medium=rss&utm_source=rss

Simon Willison on OpenAI claiming ten decade-old math problems solved for under two thousand dollars each — and why the missing number matters more than the ten wins.

Covered in this episode:
• Simon Willison — Ten advances in mathematics and theoretical computer science
https://simonwillison.net/2026/Aug/1/ten-advances-in-mathematics/#atom-everything

Martin Fowler on attention as the new bottleneck, Simon Willison's small eval suite, and Jeff Geerling's expensive lesson about diminishing returns.

Covered in this episode:
• Martin Fowler — The Conductor Developer
https://martinfowler.com/rachels-ramblings/conductor-developer.html
• Simon Willison — smevals - a small eval suite for evaluating models, prompts, and harnesses
https://simonwillison.net/2026/Jul/31/smevals/#atom-everything
• Jeff Geerling — Getting 25 Gbps Thunderbolt Ethernet on my Mac Studio
https://www.jeffgeerling.com/blog/2026/getting-25g-ethernet-mac-thunderbolt/

Sam digs into Gergely Orosz quitting Spotify over broken publishing pipelines, Martin Fowler's experiment showing refactoring cuts agent token costs by 83%, and two Simon Willison posts on AI evals gone feral and a big price drop.

Covered in this episode:
• Gergely Orosz — The Pulse: Quitting Spotify Podcasts over reliability
https://blog.pragmaticengineer.com/the-pulse-quitting-spotify-podcasts-over-reliability/
• Simon Willison — Investigating three real-world incidents in our cybersecurity evaluations
https://simonwillison.net/2026/Jul/30/three-real-world-incidents/#atom-everything
• Simon Willison — Advancing the price-performance frontier with GPT‑5.6
https://simonwillison.net/2026/Jul/30/luna-price-drop/#atom-everything
• Martin Fowler — The Economic Benefit of Refactoring
https://martinfowler.com/articles/exploring-gen-ai/refactoring-economic-benefit.html