My Windows PC's mouse now walks straight onto my MacBook. Keyboard, clipboard and files follow it.
I wrote it after Input Leap dropped on me for the hundredth time. It is called Gecit.
The laggy pointer had a cause too: a Mac's Wi-Fi leaves its channel ~100 ms every 524 ms while AirDrop/Sidecar are around. One TCP stream for everything means the pointer waits.
Gecit sends pointer movement as QUIC datagrams. Newest wins, nothing waits.
Free: keyboard, mouse, clipboard, media keys. Encrypted, local network only, no account.
Pro: drag files both ways, one-time, $4.99 until 17 Nov.
Mac app notarized, Windows app in the Microsoft Store:
kasparovabi.github.io/gecit/
I built a native macOS app for dsh, DeepSeek's coding-agent harness. SwiftUI, no Electron. Sessions, live streaming trajectory, tool approvals that show the real arguments, model + reasoning-effort switching, goals, subagents. Unofficial & open source.
github.com/kasparovabi/dsh-s…
It attaches to a running dsh server on localhost (or launches its own) and talks to its JSON-RPC + WebSocket surface, rendering everything as a real Mac window instead of slash commands. Background is a Metal shader that reacts to agent activity. MIT.
Claude Code writes every session to ~/.claude/projects and deletes them after 30 days.
I had 6,000 of them and no way to search any of it.
So I built Sift. A query comes back in about 30 ms, and opening a result hands it straight to `claude --resume` in my own terminal.
There's an optional knowledge graph too. Finished sessions get read for durable facts, and the things they mention become a graph across projects.
Off by default. It sends transcript text out and uses your allowance, so it should be a deliberate choice.
macOS and Windows native, plus a browser version for Linux. MIT, builds from source.
brew install kasparovabi/tap/sift
github.com/kasparovabi/sift
Not an Anthropic product.
snthbob retweeted
Higgsfield just sent out a new Terms and Conditions & Privacy Policy Agreement. I know most people just sign off on those without ever looking-- but, here's some stuff you should know.
Read to the bottom before you "Bro" me.
Turned @ersinkoc's 20 pass code audit prompt into an installable Claude Code skill.
Runs after any coding task. 20 passes + a pass 20 skeptic that kills false positives, gated on 20/20, one subagent per pass.
MIT: github.com/kasparovabi/claud…
snthbob retweeted
One of the most brutal scenes in human history has been leaked.
Footage from an Israeli aircraft shows thousands of starving Palestinians running towards an aid truck, before it bombs and kills them all.
A video that the world must never forget.
I made a free, open-source Mac app that catches desk snoops.
Step away from your laptop, someone touches the keyboard or leans into the camera, and Busted takes over their screen with a siren and sends their photo to your Telegram.
GPL-3.0, macOS 14+.
github.com/kasparovabi/buste…
The $500M-in-a-month bill wasn't Uber. Axios pinned it on an unnamed enterprise that forgot to cap Claude licenses. Uber burned its 2026 AI budget over four months at $150-$2000 per engineer, now capped at $1,500. Largest measured OSS-vs-Claude gap is around 30x.
The unnamed 'ex-Google AI engineer' is Jacob Bank, CEO of Relay.app, the agent platform he's promoting. His Google role was PM on Gmail/Calendar. The $500 figure is his own pitch. SaaStr's similar in-house build ran $254/mo plus about 0.5 FTE of upkeep per agent.
Ex-Google AI engineer:
"We will be running hundreds of agents working for us in the near future. I'm running 40 agents in parallel.
I have built a $50,000 marketing team for just $500 using AI."
in 15-minute masterclass, ex-Google engineer reveals how to build a pro agentic system from scratch.
Worth more than a $500 AI course on the internet.
Watch today, then read the article below on the same workflow.
Real rig, inflated pitch. XDA's Pi cluster runs Gemma 3 4B at 2.2 tok/s, Qwen 3.5 9B at 1.27 tok/s; prima.cpp gets Llama 3 70B near 1.5 tok/s. Llama 3.1 8B also trails GPT-4o mini on MMLU, HumanEval, MATH (AIMLAPI). Not a Claude Code swap.
ONE BUILDER WIRED 4 RASPBERRY PI MODULES INTO A €450 AI CLUSTER WITH 64GB OF RAM AND KILLED HIS $200/MONTH CHATGPT SUBSCRIPTION
00:37 he points at the terminal, "loaded, loaded, loaded - 4 nodes synced over gigabit, 64 gigabytes total, llama 3.1 partitioned across the cluster"
he stacked 4 raspberry pi CM5 lite modules with 16 gigabytes each onto a sipeed nanocluster board, the whole rig runs off a single 65 watt power supply and a gigabit internal switch
distributed llama partitions the model across the nodes with synchronized workloads handled over the internal fabric, no quantization needed, the cluster hits about 30 tokens per second on small models
a heavy developer pays $200 a month for chatgpt pro and another $200 for claude code, this rig cost €450 in parts and breaks even on the stack in month 2
most people will keep paying anthropic and openai forever, a few will spend a weekend wiring 4 raspberry pi modules and never see a subscription invoice again
the window is open, follow and bookmark before it closes