We now have Autohive doing its own vulnerability scanning, and patching, on itself. Not a public feature yet, but Autohive manages its own code base in sandboxes, and undertakes maintenance tasks for the team like this.

I love Amp threads. By default they are shared among those in a workspace. And I often read them to learn things. Either about bugs the team are fixing or how they are prompting to get results.

"everyone" we shipped autohive.com, already have paying organizations and thousands of signed up users. All built with AI. While it's true many AI "developers" are, in fact, influencers just getting paid musk bucks to talk about AI, people who are using it well

Summer Yue There’s a reason these destructive actions request approval on Autohive.com Love OpenClaw, but it should be seen as a fun science experiment right now.

okay so the technical backend behind latent patterns is done. time to move onto business ops automation via Autohive

Had a great morning at Victoria University today. We’ve been working with a team of 4 postgrad students on how to better automate error fixing and resolution with both Raygun and Autohive. The team did an amazing job, which resulted in identifying a multi-agent

Lee Robinson I have an agent that reads my calendar, inbox and has a perplexity tool to search about things with. I have it scheduled to run at 6am so I get an email report to read that gives me a briefing for the day, with background research, cross references meeting attendees with emails,

Homebrew is so bad.

Cooking up some sweet sweet things in Autohive

"Websockets guarantee you will hit the same server" > Sticky sessions? "You have to send the whole context back" > previous_response_id? x.com/theo/status/20…

🤯 I cannot believe that people find it acceptable for a CEO to be proud to be able to read your messages or to push bugs to increase your spend.

GPT 5.4 / GPT 5.3 Codex are good at coding, but holy shit outside of coding they are dumb as fuck. Me: "Commit the changes to the current branch" Codex: "I've created a new branch and committed them"

LGTM!

Weekend R&D project for next round of optimizations to how we deliver software Autohive & Raygun. Using Andrej Karpathy’s auto researcher, except it takes outputs from BenchmarkDotNet and then runs experiments to improve speed and reduce memory. Contrived example in the

Things that never happened with iTerm2...

The more I learn about Terraform the more I can't believe how its a steaming pile of shit.

Its a good day today.

Sometimes it feels like JetBrains TeamCity team is on autopilot, or not existent, there's so many UX improvements that need to be made, perf improvements required, but the product is stagnated.

People always talk about Firefox being shit. "Wahh gradients have banding" Yet Chrome cannot render a scaled down image using 'cover' without it being a pixelated mess. Atleast Firefox can render it retaining some quality so it doesn't look like scaled down windows paint image.

This 'Claude Mythos' model that people keep talking about from Anthropic doesn't exist. The reason they wont release it to the public is they are just building hype around themselves for a non-existent model.