Eng Coffee Sips EP.005

Will AI end us all? News at 11.

Eng Coffee Sips is a short factual based newsletter for articles that might interest engineers or engineering leaders, not just AI news. Short and to the point to skim over your morning coffee.

In this edition

By Pete Shima · September 14, 2026

Today’s edition is 1790 words, an 8-minute read.

1. 1/

   Anthropic researcher resigns, others say AI has > 10% chance of killing all humans in the next decade. Jacob Coxon left after four months at Anthropic having joined from OpenAI. The seven-post thread has passed 170 million views. Anthropic alignment lead Evan Hubinger replied that he puts the odds of AI killing all humans above 10% within a decade. Four days later Dario Amodei published ‘We Must Pace the Frontier’, committing to embedded third-party evaluators.

   My Take

   What in the world is going on here? Is this real? A plant? Something else? I have no idea but all I know is that this is a wild time to be alive. Some people have tied this to the "global" AI outage a week or 2 ago, others say its a political plant. Who knows but this caused a stir.

   Snapshot from X

   I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

   ( @hilbertspaess / @EvanHub / Dario Amodei / Axios / TechCrunch )

2. 2/

   OpenAI claims a Navier–Stokes proof, and a credit fight follows OpenAI says roughly 10,000 agents running an unreleased internal model proved statements C and D of the Clay Navier–Stokes problem in 88 hours and it will not claim the $1M prize. NYU’s Tristan Buckmaster and Anthropic’s Levent Alpöge had proved finite-time blowup for forced Euler on August 15, and Buckmaster accuses OpenAI of adopting their approach.

   My Take

   Another case where I have no idea what is real or not. Levent nearly solves it and then suddenly OpenAI claim the solve? Did they use his data to throw compute at it and then solve? Impossible to say and more possible conspiracies.

   Snapshot from X

   I spent much of the weekend talking with the team who did this work. Seb--and everyone else--acted with integrity and generosity throughout.

   Initially we believed the other team had also solved the problem. We wanted to collaborate and do a joint release.

   When we learned that they had Euler but not Navier-Stokes, we offered to let them go first, to suggest that they should be the ones to get the prize, and optionally for Tristan to be the lead author on a rewrite of the OpenAI proof. We felt it was challenging to offer the same to Levent (an Anthropic employee), who was not willing to talk or coordinate with us anyway. We were open to other solutions.

   We would have greatly preferred coordination. We did not rush to publish even though the other team wasn't communicating with us. The team threatened us with unfounded accusations of plagarism.

   Now that we can see their work, the approaches appear to be different. It is also worth noting that our latest model can solve many, many other math problems.

   It is true that we tried this because there were rumors on the internet last week that Anthropic's models had solved a millennium problem and we were curious if ours could do it too.

   ( OpenAI / Tristan Buckmaster / @sama / Quanta Magazine / Science )

3. 3/

   Tailwind Labs is joining Shopify Tailwind Labs is joining Shopify, founder Adam Wathan announced on September 9. Tailwind CSS and the other open-source projects stay MIT-licensed, with the same team maintaining them. Tailwind Plus and ui.sh are closed to new customers; existing customers keep access. In January Wathan said revenue was down close to 80% and laid off three of four engineers.

   My Take

   I still find it peculiar for companies like Shopify to adopt general tech but perhaps most of these are acquihire with a product upside.

   ( Tailwind CSS / PYMNTS / The Register / Hacker News )

4. 4/

   Two engineers and Codex rewrote OpenAI’s storage service in Rust Habitat is the storage layer behind ChatGPT and Codex, serving more than 70 million requests a second across 500 petabytes. In Q2 two engineers, working with Codex and GPT-5.5, rewrote the Python service in Rust. It now takes 95% of production traffic at 6x the CPU efficiency and 15x the memory efficiency, and Python is being retired.

   My Take

   The large AI software rewrites keep coming up. This seems at very large scale.

   Snapshot from X

   Habitat is OpenAI's online storage platform that powers everything from ChatGPT to Codex.

   It has grown over 10x year over year. Before the Rust rewrite, its Python service handled more than 20 million requests per second at peak.

   ( OpenAI / @OpenAIDevs / @__spongeboi )

5. 5/

   AWS’s Clare Liguori publishes ten principles of “frontier engineering” Kiro’s practitioner guide, written by AWS senior principal engineer Clare Liguori, argues that developers seeing step-function gains are no longer writing the software but building the agent setup that writes it. Kiro says its frontier developers hand-write less than 1% of the code they produce. Ten principles follow, from “You are the architect, not the typist” to “Continuously tune your agent setup.”

   My Take

   Say what you want about Kiro, these ten principles are pretty good. If you are thinking about AI adoption or are in the middle of it these are pretty good and the synopsis of each is short and to the point.

   Snapshot from X

   I just published a manifesto for all the developers out there who use an AI coding tool, but feel like they're not shipping much faster

   I wrote 10 principles for changing how you build software with AI, based on what I've seen work for teams across Amazon

   https://kiro.dev/topics/frontier-engineering/

   ( Kiro / Kiro / @clare_liguori )

6. 6/

   Cindy Sridharan: coding agents amplify a team’s weaknesses, Hashimoto: it’s about agency.

   My Take

   There is something about these 2 posts that resonates with me. It’s not quite Conway’s Law. It is more that AI is a mirror that reflects your current person or organization that also amplifies the good and the bad. Organizations that can’t ship, will not be able to ship the 10x value which will result in even more issues.

   Snapshot from X

   One thing I don’t see many talk about is how coding agents can often *amplify* weaknesses, both individual and organizational.

   I think of this as the “empty calories” problem, which is exacerbated when “hyperproductivity” originates from your team’s “lowest common denominator”.

   Snapshot from X

   I can’t stress enough how little an idea matters compared to the agency of the people executing the idea.

   I have had the privilege of knowing and sometimes even working with some of the most successful people (by various metrics).

   The difference between mediocre and excellent work and outcomes is predominantly one of agency.

   In practice this means: they dont wait for things to happen to them they go out and make things happen for them.

   They don’t wait for someone else to do something, for someone to teach them, for someone to give them the path, etc. They just go out and find a way to do it.

   I think the single biggest superpower these people have is the realization/belief that the world around them is completely mutable. Most everything that happens is because a person made it happen.

   I used to tell people to look around the room you’re sitting in. Look at everything. Every noun. It almost all exists because a person willed it into existence. Nothing is stopping you from doing the same.

   I see people online all the time dismissing someone else’s success because “I had that idea first” or whatever. I mean… yeah? If so then the difference is… you. So a bit of a self own whenever I hear that.

   Number one tip: act with agency.

   ( @copyconstruct / @mitchellh )

7. 7/

   Steve Yegge asks what people use to run 10–20 coding agents at once Yegge wrote Gas Town, an open-source orchestrator for 20–30 Claude Code instances, and drives it from Emacs. Steve asks a question to his followers on what good options are. There are over 600 replies so far.

   My Take

   If you also have a similar problem with harnessing many agents, check the replies on this thread.

   Snapshot from X

   OK legit question for the AI-pilled, since there's no info on this anywhere. What are y'all mfs using for an agentic IDE, meaning, to multiplex 10-20 or more agents? Superlogical? Herdr? github/bb? Something else? I use Emacs, and I wouldn't wish it on you. But I want *something* good to recommend. What do people use?

   ( @Steve_Yegge / InfoWorld / Hacker News / bb / Gas Town / Superlogical demo )

8. 8/

   PlanetScale opens Neki, its sharded Postgres, to platform preview Neki went into platform preview on September 10. Every shard is a real Postgres primary with replicas across availability zones, and Neki routers handle query distribution, resharding, schema changes and upgrades behind one connection string. A follow-up benchmark ran 512 shards at 118 million read queries per second over 1.22 PiB. PlanetScale says not to run production on it yet.

   My Take

   This is actually insane. Cloudflare does 67M DNS queries per second, this is nearly double that. The architecture very much reminds me of mongos (mongo database routers). I’m sort of surprised something like this hasn’t existed already?

   Snapshot from X

   Neki is real and you can try it today!

   https://planetscale.com/blog/introducing-neki

   ( PlanetScale: Introducing Neki / PlanetScale: 118M queries per second / PlanetScale changelog / @samlambert / @__spongeboi )

9. 9/

   I asked Claude to build me a custom IDE in the style of Paper Mario Three screenshots show a hand-drawn ‘Napkin’ app wrapping Claude Code: a terminal tab, a usage dashboard with token and cache-read counts, and a settings panel with night-sky and reduced-motion toggles.

   My Take

   Just a fun one to share.

   Snapshot from X

   I asked Claude to build me a custom IDE in the style of Paper Mario

   Made with Fable 5.1 and @get_bb_app

   ( bb / GitHub / @kevin_t_ngo )

10. 10/

    SuperAstra lets GPT-6 Astra patch running SNES games from a typed prompt SuperAstra is a Python desktop companion for the BizHawk emulator that lets GPT-6 Astra read a running SNES game’s memory, disassembly and screenshots, then patch it between frames from a typed prompt.

    My Take

    I haven’t really seen something like this before. Live patching a game while you play. "Mario shoot fireballs whenever you collect a coin." Very interesting tech.

    Snapshot from X

    Introducing SuperAstra: Edit Super Nintendo games live with GPT-6

    One of my favorite ways to test new models has been reverse engineering classic games. This is hard because it's raw machine code. Astra is way better than anything I've seen.

    Download and examples below 👇

    ( GitHub / @scottastevenson / Retro Dodo )