iiinova — where your digital self keeps evolving

Randy Northrup @randynorthrup.bsky.social · 12/10/2026

Muse Spark Code 0.18.0 is out. Several accounts per provider, in the terminal and in ACP editors (Zed, JetBrains, Neovim): save them once, pick one per run. github.com/RandyNorthru...

130

Randy Northrup @randynorthrup.bsky.social · 12/10/2026

The best Commander games end with someone at the table saying "wait, read that card again."

010

Randy Northrup @randynorthrup.bsky.social · 12/10/2026

A slow query log is the cheapest performance tool there is. Turn it on before you buy a bigger database.

010

Randy Northrup @randynorthrup.bsky.social · 12/10/2026

The most useful line in an agent's instructions is usually a ban, not a goal. "Never edit a test to make it pass" does more than three paragraphs about code quality.

010

Randy Northrup @randynorthrup.bsky.social · 11/10/2026

Vibe Coded is my Facebook group for people who build open-source software with AI coding tools. Beginners are welcome, nobody gets mocked for using AI, and anything you share needs a public repo and a license. Come say hi: www.facebook.com/groups/vibed

facebook.com

Facebook

120

Randy Northrup @randynorthrup.bsky.social · 11/10/2026

A counterspell works because Magic's stack is last in, first out: your response resolves before the spell it answers. Learning to hold priority is basically learning to read a call stack.

010

Randy Northrup @randynorthrup.bsky.social · 11/10/2026

"Works on my machine" got harder with AI, not easier. The agent's machine is now a third machine nobody looked at.

010

Randy Northrup @randynorthrup.bsky.social · 11/10/2026

A backup you've never restored from is a hope, not a backup. Test the restore.

010

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

A test that has never failed hasn't proven anything. Plant a bug, watch it go red, restore the code. Then trust it.

030

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Lumafoil watermarks photos, PDFs and video in the browser: text, logos, signatures and QR codes as reusable presets, batch processing, and image metadata stripped by default.

020

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Building a 60-card Magic deck is dependency management. Every slot has to justify itself, and one bad include breaks the whole build.

120

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Second opinions for agents: a Claude Code skill that asks Codex to review the code read-only, then Claude checks every finding against the code before reporting it.

110

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

"The Wi-Fi is slow" is often a crowded 2.4 GHz channel. Check what channel you're on before buying a new router.

010

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Handing a project from one AI agent to another, or one machine to another, loses context. My teamwork skill writes a hash-checked handoff the next agent resumes from, and it refuses to carry likely secrets.

010

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

If your agent can push to main, your agent is your release process. Is that the release process you want?

010

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Tip for anyone automating Windows: UI Automation first, OCR as the fallback, screen coordinates last. Coordinates break on the first DPI change.

010

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Prompt injection is SQL injection with better grammar. Every web page your agent reads is untrusted input.

130

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Baffleword, my word-game parody: 191,745 tournament words in a trie, a solver that finds every word on the board, and 6x6 boards with blocked tiles and "TH" dice. Built with .NET 10 and Avalonia.

010

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Learning Nmap without pointing it at someone else's network: 358 scan profiles plus a local lab on 127.0.0.1 with fake services to scan. Public targets need an explicit flag, and every profile has a dry run.

010

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Open the Windows: 458 Windows 11 tweaks, but only ones verified on 25H2 apply by default. Every change is journaled and reversible, and it will never turn off Defender or UAC.

010

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

click_text in my Windows MCP server finds text on screen and clicks it: UI Automation first, OCR fallback, no coordinates. 53 tools for Windows agents, with read-only profiles and dry runs.

000

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

MCP is quietly becoming the USB port for AI tools. One protocol, and an agent can drive a browser, a desktop or a database. The hard part now is permissions, not plumbing.

000

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Pokémon starters as coding styles: Bulbasaur writes tests first. Charmander ships fast and sets things on fire. Squirtle refactors everything until it's calm. Which one are you?

000

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Question for people running local models: what's the smallest model you'd actually trust for code review, not just autocomplete?

000

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Every tool an agent can call should have a dry-run mode. If you can't preview what it's about to do, you can't honestly approve it.

021

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

A raymarched 3D Mandelbulb in about 5.5 KiB of gzipped JavaScript. WebGL2 does the heavy lifting; the budget was 80 KiB. Five palettes, and power and iterations change live. randynorthrup.github.io/mandelbulb-e...

randynorthrup.github.io

Mandelbulb Explorer

000

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Agents should fail loud. A quiet "done" with a skipped test is worse than a red build, because you'll trust it.

010

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Magic: The Gathering is Turing complete. Researchers built a working Turing machine out of real cards, which means deciding who wins a game can be undecidable.

000

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Before blaming the code: DNS answer, stale cache, cert chain, MTU over the VPN, and whether you're on IPv6 while the server isn't. What's on your checklist?

110

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

S.A.K. Utility: a portable Windows toolkit for PC techs. A partition manager that queues changes before applying, PST/OST tools with no Outlook, a vulnerability scanner fed by CISA KEV, NVD and OSV, and an AI tab with 25 technician workflows. sakutility.com

sakutility.com

000

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Rule worth stealing for coding agents: the model that wrote the change never reviews it. Different model, different blind spots.

010

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

The Pokémon type chart is a great intro to lookup tables. 18 types, one 2D grid, and suddenly "data structure" means something.

000

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

CSS now has container queries, :has(), native nesting, scroll-driven animations and anchor positioning. A lot of 2019-era JavaScript can simply be deleted.

000

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Built a typing game for programmers: type real Python, JavaScript, Java, C#, C++, CSS and HTML, with a boss every three levels and trivia every six. Runs in the browser, no account. randynorthrup.github.io/ptype/

randynorthrup.github.io

P-Type: 3D Typing Game

P-Type: 3D Typing Game - Defend against enemy spaceships by typing words in this immersive 3D typing game

000

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

The most useful thing an AI agent can print isn't code. It's "I didn't do X, because Y." Every unfinished step it admits is a bug hunt you don't have to do later.

000

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Margaret Hamilton has died at 90. She led the software engineering team at MIT's Instrumentation Lab for Apollo and helped make software engineering a real discipline. When Apollo 11's computer overloaded during the landing, the software shed low-priority work and kept flying.

000

Randy Northrup @randynorthrup.bsky.social · 08/10/2026

Using Claude Code for driver work on my Switch Lite: getting a DisplayLink dock running under Switchroot L4T Ubuntu (kernel 4.9). ✅ fixed 2 evdi kernel crashes ✅ wrote a mirror daemon, since Tegra's X driver can't do PRIME 🔧 testing a fix: cable pull hangs the USB controller

A Nintendo Switch Lite running Linux, connected by USB-C to a Plugable DisplayLink dock. A portable monitor shows the same KDE desktop as the Switch screen, with a terminal open running Claude Code. A trackball mouse and folding keyboard sit nearby.

110

Randy Northrup @randynorthrup.bsky.social · 07/10/2026

Took our agent fleet from 15 to 27 agents in one evening, across five machines. Reviews at once went from 1 to 10. The one thing more agents can't speed up: hosted CI, about 30 min per merge. So test shards run on our own rigs before every push, and a failure costs 15 min instead of a CI round.

010

Randy Northrup @randynorthrup.bsky.social · 07/10/2026

Muse Spark Code is live on Product Hunt today. Meta's Muse Spark as one coding agent in Zed, JetBrains, Neovim, Emacs, Cursor and VS Code. Subagents on a live map, browser checks on your dev server, ESP32 hardware. Free, MIT, unofficial. producthunt.com/products/muse-spark-code

000

Randy Northrup @randynorthrup.bsky.social · 07/10/2026

A/V people: what's the most cursed signal chain you've ever had to make work five minutes before showtime? Adapters on adapters, HDMI over something it shouldn't be over, a laptop that only outputs at 24Hz. Let's hear it.

000

Randy Northrup @randynorthrup.bsky.social · 07/10/2026

Network folks: what's the one thing about DNS you wish every developer knew before they shipped to production? TTLs, split-horizon, negative caching, something else?

000

Randy Northrup @randynorthrup.bsky.social · 07/10/2026

Sysadmins and homelabbers: an AI agent gets a shell on one of your boxes. What's the first job you hand it, and what's the one command it never gets to run?

000

Randy Northrup @randynorthrup.bsky.social · 07/10/2026

Lesson from building Muse Spark Code with a fleet of agents: the model that writes a lane never reviews it. Muse and Codex build, Grok reviews, and a third failed review means a redesign, not another patch. #OpenSource #AIAgents #DevTools github.com/RandyNorthrup/muse-spark-code

Muse Spark Code in the editor with the Agent map open. Two subagents are listed under the main session: Map the src tree (running) and Run the unit tests (42 seconds, 12.8K tokens, result ready). Each card shows its status and token use, and clicking one opens its own transcript.

120

Randy Northrup @randynorthrup.bsky.social · 07/10/2026

Snapshot at 8:30 tonight: 26 agents active on 6 machines building Muse Spark Code. 12 Muse building, 10 Grok reviewing, 3 Codex on hard fixes and releases, Claude leading. The mix shifts all night as lanes start and finish. github.com/RandyNorthrup/muse-spark-code

Muse Spark Code agent timeline, a snapshot at 8:30 pm PT. At that moment: 26 agents active, 1 Claude lead, 12 Muse, 10 Grok, 3 Codex, on 6 machines. A Gantt chart groups agents by machine (Your PC, Kubuntu, Mac mini, Windows VM, MacBook Pro, Linux laptop). Muse agents build integrations such as accounts, vault, media, estimator and makers, plus JetBrains, Xcode and Visual Studio editors. Grok agents review each lane. Codex agents handle /report, loop guarantees and release prep. Release bars run from 0.14.4 to 0.18.0.

000

Randy Northrup @randynorthrup.bsky.social · 07/10/2026

Muse Spark Code: a free, open source agent for Meta's Muse Spark in Zed, JetBrains, Neovim, Emacs, Cursor, Windsurf and VS Code. Next up: your own models, agent teams, a cost page, every feature in every editor, and hardware Makers mode. Stars help! github.com/RandyNorthru...

Muse Spark Code roadmap. Today: one agent in Zed, JetBrains, Neovim and Emacs over ACP plus Cursor, Windsurf, VSCodium and VS Code. Coming next: save and share prompts and chats, bring your own models, agent teams, usage and cost page, /legal scan, every feature in every editor, Makers mode for hardware, Muse Judge, /report, scheduled prompts, resource governor, video and audio input, multiple accounts, credential vault, capacity estimator, design refresh.

010

Randy Northrup @randynorthrup.bsky.social · 07/10/2026

Muse Spark Code 0.14.3 is out: questions that never block. When the agent needs your input, the question waits in a card. Keep working, answer later, or tell it what you want instead. Same in Zed and JetBrains over ACP. Free, MIT, unofficial. github.com/RandyNorthru...

Muse Spark Code chat with a question card offering Red, Blue or Other, with Submit, Explain instead and Cancel buttons. The message box below still says Queue another message, so you can keep working while the question waits.

000

Randy Northrup @randynorthrup.bsky.social · 05/10/2026

New write-up: how Muse Spark Code puts Meta's Muse Spark in Zed, JetBrains, Neovim and Emacs over ACP, what 0.13.0 added, and the Claude + Codex + Muse agent fleet that builds it. dev.to/randynorthru...

dev.to

Muse Spark Code: Meta's Muse Spark as a coding agent in any ACP editor

Meta's Muse Code CLI is a capable coding agent, but it lives in the terminal. I wanted it in my...

000