Leigh

"People who say that crypto or blockchain does not have real utility live in a bubble, because they live in countries where the financial system works perfectly for them and because they have not seen how it is to have a broken financial system.”

@bcherny I see this in claude.com/blog/1m-contex…: > 1M context is now included in Claude Code for Max, Team, and Enterprise users...1M context previously required extra usage. But when I use 1M context, it bills extra, on both my Max and Team accounts. What am I doing wrong?

@bcherny I'm seeing odd behaviours on rollouts of new features. Across both Claude accounts I have, across multiple computers. The new features are not available. Including features that my enterprise has on, and that I've seen "on for all" tweets about. E.g. with both /rc and /voice.

Update: this is now rolled out to 100% of users

@trq212 @vivekkmkpinn I’ve noticed similar long delay on features inside claude-code when the feature is accessible outside. For example: claude --remote-control works, but /remote-control doesn’t. Is that expected?

@vivekkmkpinn yeah sorry its taking a bit of time to roll out for some people

We just added /btw to Claude Code! Use it to have side chain conversations while Claude is working.

👀 github.com/coinbase/x402/…

👀 @x402/stellar" target="_blank" rel="nofollow noopener">npmjs.com/package/@x402/…

✨ @StellarOrg sets a new all-time high A single block just reached 356 tx/s, a 40% jump in peak throughput For a network powering real-world blockchain solutions and expanding financial access, higher performance means greater reach and impact 🌍 📊 chainspect.app/chain/stellar?…

Solidity smart contracts on Stellar are looking more and more real. Excellent work by the Solang team @salaheldin_sameh/announcing-solang-compiler-suite-solidity-support-for-stellars-soroban-1fa82335101b" target="_blank" rel="nofollow noopener">medium.com/@salaheldin_sa…

This applies to most agents I think. I find myself frequently starting new threads. Often not bothering with summarising or bringing context. Use a fast model and prioritise solutions that are easy to understand and it can rediscover the relevant without the noise.

I think one of the tricks of working with AI effectively, especially if you're using a parallel workflow, is to set it up to work inside what you can take on as cognitive load. The limits of context switching and the steep effort of reviewing big prs do not disappear.

12/ Open Source & Audited Smart Accounts are live in @OpenZeppelin’s Stellar Contracts Library ✅ Open source ✅ Audited ✅ Production ready Docs 👉 docs.openzeppelin.com/stellar-contra… Code 👉 github.com/OpenZeppelin/s…

@ryan0x44 Cursor 2.0 feels like 90% a good interface for temporary worktrees that just happens to leverage that for agents. But in general we need better worktree tooling for parallel hacking. Maybe it already exists but I have not found it.

Is it just me or does the DX and docs around git worktrees feel like a lot of work? This screenshot is just the commands alone. Feels like just using regular git clone with multiple directories is an easier approach and not too different, just extra disk space and git fetches?

@tomerweller We should remove the brackets around 'epic'. It works without it. Epic go repo restructuring 💪

It's never too late to admit that monorepos are a bad idea for open source projects github.com/stellar/go/iss…

Q3 is complete and it was a big one — packed with announcements, a protocol update, and of course, Meridian. Join @DenelleDixon, @JoseFDaPonte, @tomerweller, and @rajachak75 for our quarterly webinar tomorrow at 12pm ET / 9am PT. stellar-org-us.zoom.us/webinar/regist…

@ryan0x44 @AmpCode Ok I got better results on another task when explicitly telling it to use the Oracle. Do you have an example of an effective AGENTS.md file for Amp? It seems to have some unique things, like mentioning the Oracle specifically, that I haven't needed with others.

@___leigh___ @AmpCode Do you have an AGENTS.md file set up? And for the initial analysis did you try any of the sub agents like the oracle?

With AI coding I'm spending more time discussing the approach of how code will be implemented (at a high level) now and getting almost perfect results thereafter. Which feels counter-intuitive because 5 years ago this level of planning prior to coding was slower.

@ryan0x44 @AmpCode I like that amp can see lines of code highlighted in the ide automatically and that it supports vim with that. With cursor I use a key sequence in vim to copy into the clipboard the file and line range, and paste it into cursor.

@ryan0x44 @AmpCode Both ran tests. Grok identified it needed to run additional workflows locally to update generated code.

@ryan0x44 @AmpCode Very interesting. Right now I’m using Grok Code Fast (via cursor cli) and at times I have felt the same thing. I prefer speed over perfection. Is Amp fast? I’ll try it.

@___leigh___ Right now it’s @AmpCode. With all the hype around Codex recently I gave it a try for a few days, and all I can say is people who haven’t used Amp yet have no idea what they’re missing