Build things with purpose. Open source, or for the common good.
Don't build for money. Build for the things you want to see on this planet.
Purpose → Drive → Real satisfaction.
What's your WhatsApp Claude Code workflow?
Mine: spawn terminals from WhatsApp, drive them with voice notes, native /clear /compact /context — each project has its own thread, I talk to each Claude separately.
Claude Code Pro: $20/mo.
Claude Code via byoky: $0/mo — while the token pool has gifts.
One browser extension. Your own key (or a gifted one from the pool). Same CLI.
Setup: https://t.co/2RDHtDElBi
Pool: https://t.co/FAbORvArOq
Introducing Byoky.
The AI token network where your users bring their own keys. You never touch or pay for their tokens.
→ Use Claude Max inside OpenClaw
→ Gift token budgets with hard caps
→ 13 providers, 2 lines of code
→ Zero infra — users pay the bill
More below 👇
Use your Claude Max subscription inside OpenClaw — or gift a capped token budget to someone else. Same wallet, same OAuth flow, the credential never leaves your device.
`<N> <msg>` → session N. `#tag <msg>` → session whose folder matches. `!all` → every Claude replies with one-line status. Quote-reply → routes back to the session that sent it.
Repo: https://t.co/b1sv07d3YR
Writeup: https://t.co/wcbdqiB9Ak
I wanted to run Claude Code from my phone across five projects at once. Hit three limits that nobody talks about. Each one taught me something about Claude Code's internals.
Here's what broke and how I fixed it. 🧵
@CorticalLabs Stack: Next.js 16, React 19, Tailwind 4, Three.js, TypeScript. Python for the offline pipeline.
MIT licensed. @CorticalLabs gave permission to use the recordings.
Try it: https://t.co/0oW6ElMt7n
Code: https://t.co/wU7BdH1i2b
Write-up: https://t.co/eexVz2sngp
Two cultures of real human brain cells play Pong against each other in your browser.
Every paddle move is decoded from real spike data. No simulation — actual recordings from @CorticalLabs.
Pick the brains. Tune what they 'see'. Watch them compete.
→ https://t.co/0oW6ElMt7n
@CorticalLabs Architecture: 100% client-side. No backend during gameplay.
Raw HDF5 recordings (~1GB each, with voltage traces) → Python script extracts just the spikes → 1.5MB JSON per recording → Three.js renders 59 electrodes per brain at 60fps in your browser tab.