Things I've shipped.

A 60-second flow-typing experiment with no backspace. An LLM reads your text and finds the moment your thought crystallized. That moment, in seconds, is your TTFT.

Your AI forgets you every session. Fix it with a repo. Paste one prompt, answer five questions, drop three artifacts. Ten minutes to a second brain your agent can actually read and write.

Research agents hallucinate. /research runs parallel subagents, an adversarial critic, and a citation auditor. Every claim traces or fails. Zero API keys.

Your next recruiter is an AI. Your resume should be too. A live surface for the agent-to-agent web: hiring agents hit a Schema.org JSON endpoint, humans get a chat with proof cards.

Eighteen months shipping voice AI to Fortune 500 enterprises. Three verticals, asymmetric conversion, four surprises. What I learned about demos, adoption, and the integration-surface signal.

Scope lessons from building an enterprise voice AI platform from zero. A narrow control point over full-stack. The MCP-first bet, the pivot I got wrong, the consolidation that followed.

Two kinds of competition, by design and by constraint, rest on one rule. Once you see that competition is downstream of scarcity, it stops being weather and becomes a material you can build with.

Shararat is a playful voice AI experience where two agents, Angel and Dan, debate your choices. Built for the Vapi Build Challenge to explore multi-agent interactions and creative prompting.

A guide to turning the spare laptop in your drawer into a home server that blocks ads and malware for every device on the network, hosts a family password manager, and watches your cameras. The walkthrough I wish I had when I started.

Know exactly how much Claude you have left. A tiny embedded proxy reads Anthropic's rate-limit headers live, so one terminal window shows your 5-hour and 7-day usage, pace, ETA to cap, and system telemetry (CPU, GPU, memory, disks, temps) side by side. Straight from the source, never scraped.