Dom

🤘 Founder @_houseofgiants
💻 Developer.
📝 Writing -

[Dom @Magnificode]

I'm tired of trying to cobble together several different AI brain architectures to do client work.

Client calls, discoveries, PRDs, decision logs, tickets, PRs.

There's way too much shit to keep connected when you're doing sales, PM, engineering, and marketing at the same time.

So I built something smaller:

Project Skills

Installable skills that help agents preserve the chain from meeting notes → decisions → requirements → tickets/PRs → delivery closeout.

These are the project context skills I'm using to help me run my dev shop.

Test it, break it, help me make it better.

https://t.co/OqsZRynWtg

[Dom @Magnificode]

I'm documenting how we're building with AI at my product development studio (@_HouseofGiants). There's a lot of risk involved, and I didn't want it to turn us into a content farming spam machine...

Meet Krang.

https://t.co/GL8UtikyFs

[Dom @Magnificode]

https://t.co/KlC0sbwkKl

[Dom @Magnificode]

https://t.co/KlC0sbwkKl

[Rhys @RhysSullivan]

twitter's best worst quality is that it forces all takes into an extreme

you're either a tokenmaxxing inferencel looping your agents or you're a promptchud

the truth lies somewhere in between

you absolutely should have agents that can reproduce problems in your products, look for optimizations, automatically open PRs on issues

software has never been perfect, even before AI bugs still made it to production, the more of these you can catch and can fix automatically the better

at the same time, the underlying engineering still does matter - i just spent 3 days fixing some bad product design and architecture that i originally tried to defer to agents to implement

if i wasn't looking at the code, or i was just trusting loops to fix it, my product would be stunted in quality because i would've just kept digging myself into a worse path

anyways, setup your loops but prompting your agents is also fine - get the balance

[Dom @Magnificode]

@wesbos This seems like maybe an okay way to make sure bots don’t keep training themselves on shittier and shittier AI generated slop code? Maybe?

[Aikido Security @AikidoSecurity]

⚠️ New "IronWorm" supply-chain attack: 30+ npm packages from @ asteroiddao shipped a malicious Rust binary firing on preinstall.

It sweeps 86 env vars + 20 credential files (AWS, GCP, Vault, npm, plus AI keys like Anthropic & OpenAI), hits Exodus wallets, hides behind an eBPF rootkit, and beacons over Tor. Self-propagates via npm Trusted Publishing OIDC, with backdated commits faked as claude/dependabot/renovate.

[Angie Jones @techgirl1908]

AGENTS•md IS the standard

[Wes Bos @wesbos]

the four horsemen of the apocalypse

[Dom @Magnificode]

Awesome idea and execution.

[Matt Pocock @mattpocockuk]

Preview of an AI Coding Dictionary I'm shipping later this month

AI coding sounds complex (harness, model, agent, tool etc) but it's really not. You just need to understand the terms of engagement.

[Dom @Magnificode]

The more I mess around with AI the less interested I am in ten agents doing ten different aspects of the business.

I'm way more interested I am in one that recalls enough context to be effective at helping make decisions.

[Dom @Magnificode]

@bentlegen 10x engineers in 2004 could write an HTML doctype from memory.

[Dom @Magnificode]

@Cloudflare had been making a lot of good moves lately. This is another one.

[Vite ⚡ @vite_js]

Today, VoidZero joins Cloudflare.

Vite remains MIT, vendor-neutral, and stewarded by the same wider team.
The same goes for Vitest, Rolldown, and Oxc.

Cloudflare is also committing $1M to an OSS fund to support independent development in the Vite ecosystem.

[Dom @Magnificode]

@KingBootoshi 👏

[Mitchell Hashimoto @mitchellh]

I've got an agent in a loop optimizing a renderer with the goal to minimize frame times (and tests to measure). It got times down from 88ms to 2ms and allocations down from ~150K to 500. Sounds good, right? Wrong. This is exactly why agent psychosis is a big fucking problem.

As an experiment, I rewrote the Ghostty core render state in Go, with access to identically laid out data structures as Ghostty and the exact same validation tests. I made a purposely naive renderer (simple, correct, but slow). 88ms per frame with 150,000 allocations (horrendous, lol)!

I then kickstarted a Ralph loop to bring the frame times down. I told it it can't modify input data structures or the public API or tests (they're correct), but it can do anything else it wants. It got to work.

It has worked for about 4 hours. I've spent around $350 on this experiment so far. The results?

88ms => 1.5ms
150K allocs => ~500 allocs

Incredible right? Nope.

My hand-written renderer I ported has frame times (same benchmark) of ~20us (0.020ms) and 0 allocations in the update path.

This is the problem with psychosis and lacking systems understanding. If you don't understand the system, you're going to accept that this is an incredible result. If you understand the system, you'll see better solutions immediately and can do roughly 75x better on throughput.

The people who blindly trust agent output are in the former camp. They're sheeple, overdrinking from a fountain of mediocrity.

Standard disclaimer: I use AI all the time. I like AI. The point I'm making is to not blindly accept results. Think. Analyze. Learn.

[Dom @Magnificode]

@herdrdev deserves a lot more attention, this is a very nice way to work...

[Dom @Magnificode]

Dang, in a month Anthropic fixed the internet to the point where Mythos is no longer too dangerous to release.

Glad there was no fear mongering around this at all...

[Polymarket @Polymarket]

JUST IN: Anthropic announces it will roll out Claude Mythos “in the coming weeks” despite growing fears over the model’s cyber capabilities.

[Dom @Magnificode]

@garand Nice! Let me know what your experience ends up looking like. This stuff is all so subjective. Curious to see 🤘

[Dom @Magnificode]

This is true. If you look at what Anthropic themselves do well, it’s marketing.

If you look at what they do poorly, it’s anything code related.

[Elvis @elvissun]

hot take:

as much as we all hate anthropic, opus is still the best model for marketing.

[Dom @Magnificode]

I won't disagree that out of the box if you tell both Claude and Codex to "Build thing X", Claude's will look better, but it's to the point now that the output is so clearly AI trope filled (everything is a card) that I spend the same amount of time directing Claude to design differently as I would need to spell it out for Codex.

If you leverage a DESIGN.md the point becomes moot, and Codex wins based on output. Even using a skill like Impeccable helps shape better output from Codex.

https://t.co/zxpyhwJmtk

[Dom @Magnificode]

@garand

[Dom @Magnificode]

It's pretty fascinating that GitHub (a Microsoft owned product) was hacked because VSCode (a Microsoft owned product) has such a poorly audited plugin ecosystem that one of their employees was duped into a downloading a malicious plugin...