How do people actually hit any limits on the max-tier AI plans? I run a codex remote server that is running 24/7. Vibing a project from my phone, issue triage, grilling and PRD creation for coco. Orchestration of sub-agents for slice implementation. Multiple hermes / openclaw agents for me and family with plenty of cron jobs and hooks. Code review on all my PRs. All on 5.5 xhigh or high in fast mode. I have never dropped below 40% of my weekly limit.

I am constantly improving my agent-supported coding env. Recently I started moving my env to a remote machine. All threads on all devices, long running tasks and automation. But man, seeing codex syncing and realtime threads across multiple devices really surprised me

Yesterday I challenged Codex (5.5 xhigh) to build a Cashu wallet using electrobun and coco v2. I had high expectations, but the results of a single four line prompt is really quite impressive... - Mint/Melt/Send/Receive - Multi-Unit Support - Proper lifecycle management Obviously my focus with Coco always has been making it as easy as possible to build Cashu wallet apps, but making it AI-friendly as an enabler is just as important...

I have asked about rucking a couple days ago, started immediately and so far it feels absolutely amazing. I found three old cast iron grill grids in the garage and wrapped them so my bag stays clean. Adding those 20 pounds to my daily dog walks raises HR just enough to make it an easy zone 2 workout. Zero costs, almost no added strain or fatigue, but 10-15k additional steps ins beautiful weather and nature ☀️

Matt Pocock's skills really made working with agents SO much more fun. The focus on design and architecture really brings incremental improvements with every PR while extending the agents knowledge about the domain. I added my own $pickup-issue and $orchastrate-slices for human-in-the-loop implementation

Does anyone have some experience in rucking? #asknostr

Holy cow… I am so glad I am a desk dweller by day

You might not need a gym. Have you tried building a shed?

I’m fine…

Today vs April 🤙

Coco v2 is going to be fire - custom units - custom payment methods - bolt 12 - onchain - receive via nostr - mint swaps Canary imminent

Native development is amazing these days… 1. Give codex a well defined goal 2. Enable MCP for simulator navigation and screenshots 3. Watch your app simulator hot-refresh while codex builds the new feature

🔈🤖 🧡

Join us for the monthly Cashu dev call today at 16:00 UTC (2h from now)! This is the best way to get into Cashu development and stay up to date with everything thats going on! Also there are demos 🥳 Meeting link: https://meet.fulmo.org/CashuCashuCashuCashuCashuCashuDevCall01?config=startWithVideoMuted=true&config=startWithAudioMuted=true

If you want to get together with brilliant minds and collaborate on freedom tech that will eventually change the future, you should check out @Sovereign Engineering. The application for the upcoming cohort is open, so what are you waiting for:

SEC-08 Application Form

Sovereign Engineering Cohort 08: YOLO++. Build the tools. Ship the future. Apply here. Kicking off on July 20, 2026.

I’m trying to structure my time a bit better, so I’ve decided to dedicate Thursdays to MVPs, greenfield ideas, and exploration projects. Today’s project: electrs-duress. It’s a tiny Bun proxy that sits between your wallet and its Electrum server. The proxy listens for scripthash requests and can send an alert via Telegram or webhook when a specific wallet balance is being checked. The idea is simple: configure the proxy with the xpub of your decoy wallet. If you ever find yourself in a $5 wrench attack situation, you unlock the decoy wallet. Your wallet software will naturally ask the connected Electrum server for balances, which triggers a silent alarm in electrs-duress and sends the notification of your choice. Github:

GitHub

GitHub - Egge21M/electrs-duress: A silent-alarm / duress proxy for electrum servers

A silent-alarm / duress proxy for electrum servers - Egge21M/electrs-duress

We’re back in business and getting together for our weekly Cashu TypeScript dev call today at 11:00 UTC (in about 70 minutes). Today’s call will be a review of the recent releases, as well as a quick planning and brainstorming session for the next versions of cashu-ts and coco. Meeting link: https://meet.fulmo.org/CashuTsDevs These meetings will not be recorded, and the Chatham House Rule applies.

37 weeks ago, I started working on Coco: a TypeScript framework and developer kit for building Cashu wallets and applications. The goal was simple: make Cashu app development in TypeScript as easy, modular, and reliable as possible. Coco provides a storage-agnostic core for wallet logic, proof management, mint sync, quote handling, subscriptions, operation recovery, and typed events, plus maintained adapters for Node, Bun, IndexedDB, and Expo/React Native. It also includes React hooks for building wallet UIs on top of the same core primitives. Today, after roughly 120,000 lines of code written in over 600 commits, and 55 alpha releases, I am very proud to announce that Coco v1 is now stable and available on npm under the @cashu scope. This release marks the first stable foundation for developers who want to build Cashu apps in TypeScript without rebuilding the same wallet infrastructure from scratch. But the work has just begun! Coco already takes care of the typical Cashu operations, but v2 is already being laid out. Custom units, more methods and other amazing improvements are going to land very soon (tm). Github:

GitHub

GitHub - cashubtc/coco

Contribute to cashubtc/coco development by creating an account on GitHub.

I am going to be back on @Nodesignal Podcast 🎙 - Deine Bitcoin-Frequenz today, speaking about one of the best experiences I had since I started to work in Bitcoin 🧡 This is going to be a blast 🥳

Guess when GPT5.5 was released...