Mitch Ribar

yes, thx. want a little further and added scaffolding context, full claude.md entry: ## Project scaffolding Compound-engineering outputs use a fixed repo layout: `docs/solutions/` (learnings), `docs/plans/` (dated plans), `docs/brainstorms/` (dated explorations), and `todos/NNN-slug.md` with a `status:` field. Reuse this across repos instead of reinventing it. After a solved, verified problem produces a non-trivial, reusable learning, automatically invoke the `ce-compound` skill, passing `mode:headless` as the skill argument. Only in repositories where `docs/solutions/` already exists.

@reebz Make sure you read the doc. I have an updated better prompt in it

If you're using Compound Engineering, but struggling at remembering when to run `/ce-compound`, add this to your AGENTS.md/CLAUDE.md file: "When a work session concludes with a solved, verified, non-trivial problem, automatically run the `ce-compound` skill" You can even just put this in your global ~/.claude/CLAUDE.md or ~/.codex/AGENTS.md too github.com/EveryInc/compo…

@trevin quality no argument but I find usage is lower when actively manage context window, i made a /prep-clear since i was clearing so much at 30-45% context

Despite what people say, i don't pay attention to the context window usage and don't manually manage it. Autocompaction happens and that's fine.

Fable 5 is back.

@jeremyphoward And now with this new identity verification, a reasonable claim could be Opus 4.8 is the best from Anthropic an Aussie will ever get. Will push many people to open weights (good thing).

@reebz Well that's my point. That's a big part of why they should have seen this coming. They created a huge opening. aka an Own Goal.

I disagree with this decision and I don't like it. But also... HOW DID ANTHROPIC NOT SEE THIS COMING‽ It is *the* obvious response to "this is too dangerous for anyone except us to use", since that relies on a premise ("we are uniquely good") that almost no-one agrees with.

@trevin @kieranklaassen thank you for compound engineering guys

Despite working together on Compound Engineering for over 6 months and daily messages back and forth, @kieranklaassen and I had never met in person until this week? He’s much taller than I expected 😂

does anyone actually read those long Linkedin posts littered with emojis with excessive line-breaks? genuine question.

@jeremyphoward Yes agree with you

Either Fable/Mythos is more powerful than we know or this US administration continues to punish them for snubbing Pentagon

Skimming an AI answer cost me 100 passwords open.substack.com/pub/mitchribar…

@LonnyLot @trevin github.com/everyinc/compo…

@trevin Is there a link somewhere to the repo? I didn’t see it in your profile…

glad to have chipped in a tiny little bit of help to the best workflow plugin going around ❤️

This is a super exciting release - Claude Fable 5 is the same underlying model as Mythos but with added safeguards. The benchmarks are great and it's SOTA on everything by a margin but I'll add that *qualitatively* also, this is a major-version-bump-deserving step change forward (imo of the same order as Claude 4.5 was in November), peaking especially for long problem-solving sessions on very difficult problems. You can give it a lot more ambitious tasks than what you're used to, the model "gets it" and it will just go, and it's never felt this tempting to stop looking at the code at all (but don't do this in prod!). The model still has quirks that people will run into and the safeguards are configured to be a little too trigger happy for launch, which can hopefully be tuned over time. I feel a lot of things changing as working software increasingly comes out on a tap. The Jevon's paradox kicks in and I feel my own demand for software growing substantially. You can ask for anything - explainers, visualizers, dashboards, bespoke single-use apps (e.g. a full wandb that is hyper-specific just for your project), you can 10X your test suite, auto-optimize code, run giant research projects with custom HTML for the results, anything! "Free your mind" (Matrix ref). Really looking forward to all the things people build!

@jeremyphoward @dpetrou @karpathy Would you agree that Anthropic is best org to hold the keys on model like this? If it were another company, they wouldn’t have deployed with safeguards and consideration

@dpetrou @karpathy Yes. Locking in a permanent status quo power structure. Incredibly unsafe, and damaging for humanity's prospects.

@karpathy This is not a day for celebrating, Andrej. It's a very dark and very sad day, and the damage may be impossible to undo.

@emonuxui yep - move fast and break things sounds similar, but the outcomes are radically different

@reebz The idea of making errors cheap is what actually unlocks momentum. People don’t freeze from uncertainty, they freeze from irreversibility.

I don't believe in big decisions. The ones that look enormous are usually a long chain of micro-choices that we bundle into a single moment after the fact. One big brave moment? Nah. Maybe in Hollywood or AI-slopped LinkedIn posts. This is hard to believe, so try it by walking it back from something big you chose or did. There would have been a set of micro-choices leading up to it (going to the place or party, booking the taxi, talking it over with friends before you act, and so on). Before that you had already been thinking it over for a while. Earlier still, you had cleared a bit of room in your week just to make space to think about it (this might even be losing sleep over the thing). TLDR: there is always an earlier step, which means the outcome got built a little at a time rather than gambled in one sitting. @AnnieDuke's Thinking in Bets is the best practical guide I have read on this. Treat decisions as a run of small bets under the pressure of uncertainty, and never judge the decision on the result... that is the most important part. I used to play a lot of poker, which made it easy to pick up other books on this topic, but Annie's book is truly for anyone. With small bets you can limit your downside, so mistakes are cheaper to make. @shaneparrish puts it well* (I'm paraphrasing): the real test of a decision is how cheaply you can be wrong about it. When an error is expensive, people freeze and defend the call they already made. Make the error cheap and the same people start moving, testing, and adjusting as they go. So to everyone working in customer decisioning, customer engagement, customer value management, customer lifecycle marketing, or customer growth... whatever label the corporate world can't agree on: you are responsible for LTV, and interventions or treatments will come and go. All of it is a long run of small, recoverable bets that compound into lifetime value. In practice that changes how you run the work. The temptation is to save up for the big swing, the quarter-defining campaign or the model rebuild, then launch it rarely and defend it once it is live. The teams that compound do the opposite: they put lots of small treatments into market, keep each one cheap enough to be wrong about, and read the result honestly. That last part is where most of the edge sits, because a treatment is only worth keeping if it actually delivered a result, rather than riding the coat-tails of a result already in motion (the age old difference between correlation and cause). Cut the losers early, let the winners run, and the wins stack up into LTV the same way the micro-choices stacked up into the decision. The same is true of you, outside of work. The life you want gets built the same way, from a long sequence of small, cheap, reversible moves, most of which you can walk back if they do not land. Lived like that, the whole thing feels lighter and a little braver, because you can start before you are sure and correct as you go. alwaysdraft.com/i-dont-believe… *fs.blog/brain-food/aug… alwaysdraft.com/i-dont-believe…

@garrytan @businessbarista going even further from factories to towns: gas town/gas city by Yegge... are you a fan?

@businessbarista I love a software factory but… agents calling code is more powerful

In 2023, everyone was hype about ChatGPT. In 2024, it was GenAI. 2025 was the year of Agents. And 2026 started with OpenClaw, but now attention has turned to The Software Factory. Unless you're an engineer or take residence in the depths of X, you may not know what a Software Factory is or why you should care. But when some companies are attributing 90% of their production software to AI (read: Anthropic) and best-in-class ICs are matching the output of a 20-person pre-AI engineering org, you need to care. So let me break the whole thing down... What a software factory actually is, why it's suddenly everywhere, and a simple way to figure out exactly how close your org is. Even if you've never written a line of code in your life.