Devs with JS/TS/JSX/TSX repos, this is for you. Vibe coders: coming from Lovable, Bolt, Replit, or V0? Get deep understanding of what every file in your codebase contributes to your project. If you care about deeply understanding your codebase, this is for you.

Open-source is strained right now and one of the problems is code review bandwidth. So Dana Fine and I partnered with Google AI on how maintainers can use free AI code reviews to keep up with AI-accelerated contribution velocity. With Qodo wired into PRs, you get: -

Exciting Monday... Woke up to enterprise inquiries from the other side of the world. Good start! #provablysafe #aiengineering

The back-of-the-hand math is that for a large enterprise engineering org, it could cost $~50k a week to run this in production. Imagine 2,000 PRs (especially with more AI-generated code now) going out a week. 2,000 x $15 = $30,000/week $30,000 x 52 = $~1.5M/yr I just had

"Claude, hit this 9 iron for me real quick..."

It's time to talk frankly about this concept of simply "throwing more compute at coding agents" as a means to achieving code quality.

Working with Gemini is like working with an overconfident coworker that just got kicked in the head by a horse but doesn't know it yet. Even on the first reply, which is even more maddening. My watch keeps thinking i'm having a rage stroke from some of the responses.

The Evōk Semantic Coding Engine Understanding Layer (File View) This is what you see in Codebase.Observer for a given file. Plus a Copy for LLM button to provide conversational context to your chats.

Have a react site you want us to analyze? Reach out. Private beta open for TS/JS/JSX/TSX

I especially like when I ingest the Evōk Semantic Coding Engine codebase into Evōk itself inception-style. Today, it saved me 20 minutes waiting for what would have been a failed deploy on a minor case mismatch I didn't even notice.

Finally tried Copilot CLI - I think it's rather good. Their multi agent worked well for a few tasks I gave it. This still isn't the final form, CLIs merely an access bandaid to build on, but it's all amazingly cool if you think about where we were only a few years ago.

The absurd rate at which Copilot subsidizes these calls blows my mind. Had I let the agent run this same thing on my API key, I would have spent real dollars at $3 per mil in, $15 per mil out. Instead im paying like 36 cents for this. I feel like so many of the tools developed

The spam bots are annoying and all here, but some of the usernames are pretty creative. Just got a message from LezBeeAnne for example. I would have gone for it but I'm over my quota on 33 year old Pakistani dudes.

What if codebase errors just surface automatically, as part of a repo simply existing? Not heuristics. Not LLM guesses. Mathematically provable structural, contract, and temporal errors, just surfaced for you.

The people talking about "Mixture of Agents" as some genius thing, I don't get it. I was genuinely hoping this was some change to underlying architecture, but nope... So far everyone explains it more or less the same: "Your query goes to 3+ agents at once. They all reply. Then

"Update the log status for v4, and kill that dead function. It's pointless cruft now." "IF YOU ARE CONSIDERING SELF HARM, PLEASE CALL..." Holy shit I wasn't but with that dumb interpretation I fucking might be.

I like X because I can say anything and absolutely no one will see it.

Javascript is better than Typescript Fight me.

Tech supporting dogs and cats? Instant repost.

I'm always working on 2 or more projects. Definitely the adhd, but I've also been operating this way for decades now so I know no other way. I've found it's how I stay engaged in problem-solving properly, helping avoid myopic viewpoints or missing the forest from the trees.