If you are a Senior+ engineer today and you still strongly believe you can outperform agents delivering reliable and safe code. I have some really bad news for you.

Applications are now open for the Anthropic Residency, a paid, fixed-term research program for academics, technical founders, and staff-level engineers who want to make AI safe and beneficial for everyone. No ML background is required.

Apply by November 22: alignment.anthropic.com/2026/anthropic…

one week in my takeaway is that you don't need unlimited tokens to do great work, the $200 plans are amazing

and this is the most expensive intelligence will ever be

Ok, tried out Rex by @mitchellh and @superlogical - it's fast and smooth - for a beta this is very good quality!

San Francisco traffic is crazy - 95 minutes to go across the bridge into the city and I’m not even at my destination…,

What’s up GitHub universe! Come hear me talk at 10! Say hello! We’re talking about not reading the code anymore, in choose your own adventure style. Two presenters, neither have read each others slides. It’ll be the funnest talk at the show.

Last week @nickstinemates took me from a fresh Swamp install to a 4am cron job that triages the neglected issues on the Rawkode Academy GitHub.

The agent works it out once, writes a TypeScript extension, and from then on it's just a tool call. Same result every run.

Full session: rawkode.academy/watch/hands-on…

Brutal truth from @Lethain

“If you're in [middle] management (…) you sort of have to decide, am I retiring now or am I re-skilling?”

He did just this: step down from leading a large team (250 ppl) to be hands-on with a small team (25 ppl), ship code & re-learn the new “rules”

Scanning and patching CVEs is, as everyone knows today, a real pain. Here's @sfalcon showing you how to do security scanning for the PHP vulnerability with Swamp: youtube.com/watch?v=NuxZGE…

Waiting patiently for @deno_land to sort the merge conflict out for Deno 2.9.8 so that we can roll out the fixes for the eMudhra cert authority

I am thrilled to tell you that Swamp Club has raised $7m in seed funding from @AmplifyPartners, Encoded Ventures, @jesserobbins, and @stolinski. Swamp is a platform for building AI systems from repeatable behaviors, rather than from token-hungry agent-centric designs. More money (which I'm stoked about,) or the product itself (which continues to surprise and delight me every day,) is the incredible group of people who have started solving real, practical, everyday problems with Swamp. They teach me new and exciting things about how to build with AI every day.

I'm grateful to my co-founders, who have poured every ounce of their creativity, passion, and intelligence into Swamp. I'm grateful to our users and customers (big and small) - there is no greater gift you can give to an entrepreneur than the validation that comes with voting with your wallets. And I'm grateful to our investors, who see the opportunity we see, and want to make it happen with us. It's such a cliche to say that there has never been a better time to build - but really, it feels like there has never been a better time to build. Come join the Club.

swamp-club.com/press/swamp-us…

We're throwing a little meetup and getting together to talk about Swamp / the cool things people are building with Swamp in San Francisco on October 28th.

swamp-club.com/meetup

Come meet the team, listen to some talks, talk with other community members, and eat some good food.

Here’s the truth about where the industry is at with AI adoption: most of it is awful, and most of us are suffering under the weight of it.

Early on, the bet was that the models are really smart and we will just chuck everything at them and let it sort it all out. This has largely led to incredibly wasteful, in-reproducible solutions - like huge skills and MCP, where the agent re-solves your problem a million times, getting it right some percentage often. Turns out it’s hard to get that to produce real value beyond individual experiences.

As engineers, we have struggled to get our heads around how to use these capabilities effectively. Is it the harness? An agent? A factory? A tool call? Maybe a skill? How about a dark factory! We need RAG! No, RAG sucks, we need… you get my point. Everyone needs to take a breath, and start thinking about this new capability as a component of your systems, rather than the substrate upon which your systems get rebuilt.

Meanwhile, coding agents got fucking fantastic. But for most of us, we changed nothing about our SDLC except that now we should be doing it all faster, because agents. That leads us to be stuck like a surfer in the chop - being responsible for everything but in control of nothing, with an avalanche of code nobody looked at when they wrote it, but that you’re responsible for ensuring doesn’t break the world.

Luckily, we can stop the madness spiral. Breathe. We can engineer our way out of this problem.

Are you working on how to use AI to improve your SDLC? If so, start by looking at how it works now, and using AI to automate it. Where in the process do you need intelligence? Is it there. How much do you need? Use that to decide your model. Encode the standards.

Swamp, and primitives like it, are a part of how you do that. They solve problems like how to break the problem down into reusable components, how to store data, how to build the contract between components, how to string them into workflows, and how to expose that to agents, ui, or whatever you want. The way out of the suffering is to turn our engineering brains back on, and start building in ways that enable and support the future.

Hey @AnthropicAI we still waiting on you to fix that broken GitHub Action build so people can use Opus 5.5 in CI

github.com/anthropics/cla…

IT JUST GOT INTERESTING!!! As a person who didn't find a model I liked past Opus 4.6 (except GPT Terra...)

Now I have my choice :D

As if by magic, the normal message of "this new model is great and it has been my daily driver for the last few weeks" - that means the day they shipped Fable 5.2 they changed away from using that model and used this instead.... curious

Been trying to upload a GitHub Release for about an hour now with no error but it's doing nothing - @githubstatus says there's no issue but their API says otherwise...

Had a really fun time talking at @kc_dc today - had so many fantastic questions and enjoyed hearing people’s views on building with AI

I will post the slides on this page this evening: swamp-club.com/kcdc

I think I am being dumb - but there's really no built in equivalent of `claude -w` for @opencode ?

So the majority of coding agents have standardised on AGENT.md and the same skills style - but the way to configure your own settings in them differs wildly :/