Johann Hipp

@yacineMTB @0xSero What about someone's willingness to try new things out? Lots of people do very consistent work but have an innate unwillingness to try out new workflows. Those who never try out new harnesses, skills, different models etc. There's a HUGE mindset component to this.

You get good at anything by doing it a lot. Just do things a lot It's not time spent. It's volume of doing. Make sure the time it takes you to do a single thing is as fast as possible. Your cycle time is sacred

Huge thanks to @gregosuri and @lukaspostulka for stopping by and judging the submissions :)

If that's you definitely check out weave (ataraxy-labs.github.io/weave/), which uses some oddly simple entity-level semantic merging. Has saved me a few hrs in the last week alone. Huge shoutout to @rs545837 and @Palanikannan_M for working on such foundational tools :)

Agents are working on increasingly complex PRs concurrently, so merge conflicts can end up being a huge mess since these same agents don't really understand other agents PRs.

Some tools we (or out agents) use daily are pretty bad, yet we continue using them. git has some horrible merging strategies that are barely useful for the simplest of cases and require manual rebasing/merging for pretty much anything else.

@JustusSchoch @louismorgner @garrytan @useopencompany 😭

@louismorgner @garrytan @useopencompany @jhipp_ 🫡 aight

OpenClaw is amazing until it does something funny like alter its own config file because you asked to try the new dream mode and then it can’t restart because of invalid config with the wrong version. This happened to me yesterday morning. I had to go ssh in with Claude Code to get it fixed. But the sheer fact all of these can happen shows you we are at Apple I stage: assemble your own motherboard stage. But the Apple II personal computer moment (where anyone can go to the store and buy a PC and it works) is coming for OpenClaw We are so early and this is the worst it will ever be

for @useopencompany, we're likely going to develop a native implementation. Not because it's fun, but out of necessity :)

It's going to keep on being a takedown cat-and-mouse game between anthropic and devs, but as more niche repos are maintaining Oauth hacks (most of which are exclusively vibecoded) I can see these eventually dying since most people can't be bothered and will likely just go back to claude code.

Turns out the tools most of us use to circumvent claude's vendor lock-in (vibeproxy etc.) mess up the simplest things :(

Applying this winning mindset at @useopencompany as we speak

kind of weird to someone who has never built agent harnesses before, but I fully get why some of the best harnesses' prompts are 100s of LoC long src: github.com/elder-plinius/…

frontier models are surprisingly good at figuring out how to use subagents effectively (by just having their tool definitions in the system prompt) yet will absolutely refuse to use integrations in a similar setting unless very verbosely told to do so :(

.@useopencompany gets subagents :)

shoutout to: oauth workarounds: github.com/router-for-me/… a nicer GUI: github.com/automazeio/vib… and ofc: ampcode.com

src: github.com/johannhipp/amp…

Today I'm releasing amp-proxy, a recent sideproject that I use daily. It's a tiny golang API to make amp work with claude max, codex plus, and many other subscriptions. Auth your subs, set up some model re-routing for models that amp uses but you don't have - and you're good to go :)