Two Claude Code sessions, one task list, and the same task number twice
Two AI coding sessions ended up with T-408 for different tasks within an hour. Why last-number-plus-one breaks, and the prompt and two grep commands I use now.
Tutorials
That's a Wrap on Hacktoberfest 2026! 🎃👻
Hacktoberfest 2026 wrapped up on October 31st! Thank you to everyone who built, hosted, mentored,...
How to Switch AI Character Chat Apps Without Losing Your Chats, Personas and Lore
Switching AI character chat apps is common: features change, free tiers shrink, rules shift, or...
Ctrl+C in a Coding Agent Is Just Byte 0x03. Raw Mode Decides What Happens Next
Agent TUIs can put the terminal in raw mode, so Ctrl+C arrives as byte 0x03, not SIGINT. Here's what that means for Esc, exit and the tests they spawn.
Anthropic FlashProve Shows How AI Outputs Could Become Cryptographically Verifiable
Anthropic has introduced FlashProve, an open-source research prototype designed to cryptographically...
NATS JetStream vs Apache Kafka vs Apache Pulsar: Latency SLAs, Throughput & Multi-Tenancy
NATS JetStream vs Apache Kafka vs Apache Pulsar: Latency SLAs, Throughput &...
The Biggest Problem With AI Test Automation Isn't Writing Tests
A few years ago, creating automated tests was one of the biggest bottlenecks in software...
What I Learned Managing a Full Project with Antigravity
How my workflow evolved from solo prompting in AI Studio to managing a multi-model AI team, running...
Stop Letting Five Coding Agents Ping You. Answer Them in Batches From One Inbox
Per-agent pings turn five coding agents into five interruptions. Turn them off, let blocked agents queue, and clear them at your breakpoints, oldest first.
OpenOS Momo AI — a local, offline compression AI that speaks (DE/EN), inviting independent verification
Hi DEV! I'm Sebastian (sebklaey), and I built OpenOS Momo AI — a private, local, offline-capable PWA...
Building a Federated Copilot: Part 1, Identity Before Integration
Design the human, agent and delegation contract, then prove it in a local simulation before connecting real identity providers.
I'm sure there's lot's of other weird trends AI loves to create, but this one is quite funny
Wikipedia dropped this UUID in 2014. Now it's everywhere ...
Giving an AI agent eyes on a running React Native app
Over the last few months I've posted small findings about AI agents, MCP and React Native internals....
I burned a week of Claude quota in 3 days. Then I found the setting eating it.
I burned a full week of my Claude Ultra quota in 3 days. My first instinct was to blame the model....
The Registry That Wasn't Enough: How We Went from "Knowing APIs" to Actually Verifying Against PyPI
The Registry That Wasn't Enough: How We Went from "Knowing APIs" to Actually Verifying Against...
Why freezing completed Markdown blocks reduces streaming rendering work
An AI response arrives as text deltas. A Markdown renderer normally receives a document. Connecting...
I measured whether my Claude skills help. One of them made the code worse
A SKILL.md file is a program written in English. The model runs it, and like any program it can have...
GDDR7 vs HBM3E for Inference: Bandwidth per Dollar Stops Being the Question
GDDR7 delivers 1.5 TB/s at a third of the cost. For batch-one serving that is enough, and the reason most inference clusters still buy HBM is not bandwidth at all.
DPU Offload vs Host Cores: When the 100G NIC Stops Being Free
Every inference node now spends 2 to 6 cores just moving bytes. A DPU costs $2,000 and buys those cores back, but only above a utilisation threshold most teams never measure.
InfiniBand vs Ethernet for AI Fabrics: The Congestion Bill Arrives Later
Ethernet won every market except the one that matters most for training. The gap is not bandwidth, it is what happens in the 200 microseconds after a link saturates.
Durable Execution for AI Agents: State, Checkpointing, Retries and Human-in-the-Loop
How to build AI agents that survive failures, resume interrupted tasks, handle approvals, and...
8 jq Commands That Show What Your Coding Agent Actually Did, From Its Transcript
Claude Code and Codex log every session to JSONL on your disk. Eight jq commands pull out the shell commands, edits, errors, denials, slow calls and model.
We Measured 1,500 Lines of Terminal Bloat Burning $0.20 Per Prompt in Cursor. Here's How We Stripped 97% of It in 2.5ms
How we engineered termlens: a zero-token terminal compactor and semantic error lens that eliminates 97% of CLI noise in
A Self-Hosted Code Review Bot Invented a CVE That Was Never Issued
TL;DR A code review bot is a program that reads a proposed code change and writes a...