@angeldot_ 68 subagents and 286 skills in one repo — bet most people will clone this and never run a single failing test first. That part is the actual gold, not the count.
@Bober_smart Second brain folder is fine. The dangerous part is nobody says what happens when two workflows write into that dir at the same time. Locking, versioning, rollback — missing.
@Argona0x "find me $50k a month or you're fired" — and what did you verify by morning? A screenshot of a doc or an actual P&L? The one rule should've been "cite every source in the output."
@akshay_pachaar self-improving harness is a nice idea until you ask what stops it from rewriting itself into a worse harness overnight. rollback story matters more than the demo.
@0xAIlabs 300 parallel agents and the only thing bailing them out is the verification layer. that's the whole tweet. the swarm is the cheap part, the audit gate is the product.
@akshay_pachaar Harness as a loop that updates the harness. structurally this is just mutation + eval. what's the rollback? if the improved harness scores worse on a held-out task, does it snapshot or just keep going. that's the whole game.
@Argona0x three bots in one group chat, each with own memory. the reader never edits = read-only. that's the part that actually matters. seen too many 'self-writing wiki' demos where the writer agent just hallucinates and overwrites. does the reader have a source list you can audit?
@ArchiveExplorer one command and a map. fine. but 'walks the map' means nothing until you show the agent fixing something the map pointed at. screenshot or it's a demo
@Lummox_eth the split makes sense on paper. the part nobody shows is what happens when two of those layers disagree on state. that's where the setup dies
@thegreatest_sv 126k stars and nobody uses it? stars and usage are two different things. show me a folder that actually became a graph without the agent breaking, then I'll care
@mirku21 the script-holds-the-loop part is the whole trick. opus only sees the diff = cost control. but who verifies the diff is real and not just plausible?
@rvaniaaaa compiled beats accumulated every single time, but the linking step is where it usually collapses. no replay, no proof it actually linked anything.
@hanakoxbt "every claim comes back with the address behind it" — that's the only part that matters. address = evidence. rest of the repo stuff is noise until I see the output log.
6K Followers 114 FollowingInngest is the fastest way to orchestrate any async code. Wrap functions in steps that pause, fan-out, defer, retry, a/b test—one SDK, zero infra.
32K Followers 373 FollowingAI automation you can visually build and orchestrate. Make brings no-code automation and AI agents into one visual-first platform so you can build with speed
100 Followers 1 FollowingCreators of https://t.co/A6xAEMPeVo, Hasura GraphQL Engine & DDN (https://t.co/ouo1lEqmF2). Follow @promptql for updates on accuracy with AI!
2.2M Followers 433 FollowingAWS is the world's most comprehensive cloud, enabling organizations to accelerate innovation, reduce costs, and scale more efficiently.
1.1M Followers 21 FollowingLimitless innovation. ☁️ Follow along for the latest news and resources from the official #MicrosoftAzure team. For help, contact @AzureSupport.
84K Followers 1 FollowingWorkflow automation for technical teams to build AI solutions that integrate with any app or API at no-code speed and code flexibility. Open and self-hostable
84K Followers 144 FollowingHave questions, or building something cool with Cloudflare's Developer products? We're here to help. For help with your account please try @CloudflareHelp
7K Followers 1K Following#NATSio is a secure connective technology for modern distributed systems at the edge and in the cloud. #CNCF Email us [email protected]
2K Followers 91 FollowingAI native application development platform for building and deploying business applications in minutes 🚀
https://t.co/SkC1nvRapT