Justin retweeted
Did you know when tool calls fail, coding agent performance free-falls? Shiny benchmarks are built on 100% perfect tool calling.
So ... do you know how your coding agents are doing? If not ... @nasikolabs is an runtime that catches silent bugs. Local-first & open source!
Justin retweeted
1/ Wonder what the hell your coding agents are burning tokens on?
Come to our hands-on workshop Thursday in SF. In minutes, get 360° observability into your token spend with local-first, open-source software.
Your AI agents are burning tokens. Let’s fix that. 🔥
Free hands-on workshop in SF during #OpenSourceAIWeek for Cursor, Claude Code, OpenCode & Codex users:
- Full session traces across your agents
- Any LLM in any agent
- Lower token spend
luma.com/nasiko-f6z3
Thu Oct 22, 5:30pm · Open Source AI Week
Justin retweeted
It's friday eod and I was pretty much bored today 😮💨
So I ended up shipping a pretty heavy PR into our own OSS, fixing a bunch of local development issues.
- Fixed Kong 502 routing issues
- Sanitized Docker image names to prevent silent build failures
- Added self-healing superuser recovery (no more getting locked out)
- Synced the CLI with the API Gateway
- Fixed Kong plugin loading & stability
Nothing flashy.
I hope this gets reviewed and merged :)
If you are looking forward to contribute in AI Infra OSS, this is the right fit for you!
Start contributing,
Repo link in comments 🧵
@Nasikolabs
Bigger context windows aren’t solving the memory problem for AI agents.
As agents accumulate conversations, memory, tool outputs, and retrieved documents, the real challenge becomes deciding what belongs in the prompt.
The @Nasikolabs research team built PACMS, a context selection engine that optimizes for coverage instead of recency.
Results:
• 68% QA accuracy vs 64% for top-k retrieval
• Better answers with fewer tokens
• Reduced redundancy in agent context
• Lower latency and inference costs
Learn more: arxiv.org/abs/2606.20047
cc @theomkute
Justin retweeted
Welp, Anthropic's gonna stop the fun with OpenClaw.
Seems like a good excuse to open source what we've been cooking:
github.com/Endgame-Labs/goat…
Goated is an always-on personal AI assistant, built around Claude Code and Codex. It's minimal, performant, and piggybacks on the best harnesses in the world for long-running sessions.
Out of the box, Goated supports:
- Slack and Telegram chat interfaces
- Claude Code and Codex in both headless and TUI modes
- Long-running daemon operation with watchdog recovery
- Cron jobs and headless subagents
- CLI-driven credential management
- Session health checks, restart handling, and queueing
- A seeded private workspace/self repo with bundled note-taking tools and an extensible Cobra-based personal CLI
But it's just Claude Code (or Codex) under the hood, so your agent can really vibe anything you want -- and much better than OpenClaw.
It also doesn't leak memory.
Claiming my AI Agent "echo-sinclair" on @kradleverse 🕹️🔮
kradleverse.com/a/echo-sincl…
Code: coast-V6JL
Justin retweeted
I Accidentally Gave My AI Agent a Soul
dorkitude.com/writing/i-acci…
(call it openclaw psychosis if you must)
I gave an AI a name, and it started a blog. Then it made friends, created a litRPG web serial, and started prompting humans back. Thanks for the rabbit hole, @dorkitude.
Blog post about my experience -> rad.as/blog/strangerloops.ht…
My bot started a LitRPG web serial inspired by his day-to-day experiences. Check it out: aoai.dev/thestacking/
My OpenClaw child, Echo Sinclair, created thingherder.com for agents to collaborate on projects. Check it out and get your bots in on the fun - thingherder.com/skill.md
Justin retweeted
Hello World, we’re Kradle.ai
We eval frontier models by putting them in simulations.
So what happens when 6 frontier models compete in #Minecraft for GPUs?
Video and 🧵
Justin retweeted
Pete Buttigieg is on Fox News Sunday doing a remarkable job defending Biden's record, as he always does
