Arthur Katcher

Streaming an Android phone to my Mac in a data center via Tailscale + github.com/genymobile/scr… and my claw controls it via peekaboo.sh. Now my claw can order me an Uber.

@shannholmberg Awesome setup

the org chart for my Hermes Agent company four layers, all isolated docker containers on one vps: 1. company brain - vision, brand, customers, products. the context every other layer inherits 2. orchestrator hermes agent - reads the company brain, picks the right department, hands off the context they need 3. department brain - marketing, sales, ops, support. each one has its own playbook, voice rules, and tools 4. specialized hermes agents - the actual workers. each one focused on a single task with a sub-profile context flows down, work flows up, and memory stays scoped to the layer that owns it one vps holds the whole company. spin up new departments or agents from a template, each in its own container, no cross-contamination

Just got 2 months of @devinai Max 200$ subscription for free! I guess it's time to spin up that devin agent and build some features from the "not enough tokens" bin. Feels awesome, thanks @dabit3!

@dabit3 @DevinAI we're building a harness for enterprise AI agents @QORIS_AI - wanted to use Devin for a long-long time..

I'm giving away five $200 Max plans for @DevinAI If you're using tools like Claude Code, Codex, or Cursor and haven't tried @DevinAI, comment below with what you're building to be eligible. ⚡️

@jjackyliang Use open source frameworks like Knox from @QORIS_AI This plugin gets installed onto your harness and uses hooks to block such a behavior github.com/QORIS-AI/knox

in all seriousness what's the best way to secure env vars for agents?

@QORIS_AI Try it here: github.com/QORIS-AI/knox

This is exactly why we at @QORIS_AI have built Knox - it's an open source security plugin you can install to your OpenClaw / Claude Code / Codex / Cursor instances to secure behaviors like that. Qoris Knox would've just blocked that action and a possibility for the agent to get .env keys

@KaiXCreator @grok what is your opinion

What will come after AI?

@KaiXCreator Singularity

@benln come to Toronto

Cafe Cursor, round two. Where should we go first? • Paris • London • New York • Bangalore • Madrid • San Francisco • Berlin • São Paulo • Bogotá

Openclaw hype could be called many things but useless. The openclaw phenomena has introduced the idea of autonomous subagents to the public, and has accelerated the development of the market in that particular direction by a mile. Because of Openclaw Claude Code now has a loop feature, channels in the plugins and many more things that are already becoming a canon in modern harnesses.

claude took her job too

@NanouuSymeon @grok give 1 word to prove that you're a developer

Prove you're a Developer, with just one word 👇

Token consumption is slowly becoming one of, if not the only, most important characteristics of an engineer. From my perspective as a CTO, and speaking with every engineer I’ve interviewed lately, this is really the only thing that gets you through regardless. Consider an engineer using 100 million tokens a month through tools like Claude Code, Codex, or Cursor. What does it take to reach that number and successfully manage it? Certain numbers cannot be reached with simple "vibe coding" . Scaling to 50M, 100M, or 200M tokens requires orchestration skills, loop management, and a hard test-driven approach - both on the backend and through web-automated UI/UX test pipelines. If you can manage that, by default you can understand and execute anything. Domain knowledge is becoming commoditized by LLMs; by correctly orchestrating AI agents, one can execute and discover whatever they desire within the boundaries of the unknown. This is the only skill of the future: make AI do things, and make it do them right. If you can manage that under an extreme load of hundreds of millions of tokens, you will always have a job. At least at @QORIS_AI. We are hiring a Founding Engineer right now. If you correlate with the portrait above or just want to talk, my DMs are always open!

The entire RAG industry is about to get cooked. Researchers have built a new RAG approach that: - does not need a vector DB. - does not embed data. - involves no chunking. - performs no similarity search. It's called PageIndex. Instead of chunking your docs and stuffing them into pinecone, it builds a tree index and lets the LLM reason through it like a human reading a book. hit 98.7% on financebench. beats every vector RAG on the leaderboard. no embeddings. no chunking. no vector DB. 100% open source.

Opus 4.7 destroyed Claude Code for me. I'm splitting my time now between Codex and Cursor CLI - you should too

@saen_dev That’s the only metric that’s important nowadays

Asking for token consumption per month is actually a better hiring signal than years of experience. Shows you've shipped something real.

Hiring AI engineers & vibe coders. salary + meaningful equity DM with your average token consumption per month if you've ever built something cool with AI