The best engineers blur the lines between frontend, backend, and infrastructure. With AI systems, this mindset is essential. LLMs introduce new failure modes. You need structured retries and feedback loops that explain behaviour. Ship fast, own outcomes. sailhouse.dev/blog/enginee...
I’d say most of my work recently has had part done by Claude code.
Often will have a bunch of git worktrees set up working on different things.
- We invested heavily in CLAUDE.md
- The feedback loop of tests has been so helpful to it
- It still sucks at design, but can one-shot a lot
The workflow is dead. Long live the agent! 🤖
We're shackling intelligent AI agents to 20-year-old orchestrators. The breakthrough: workflow IS the code. Agents should decide their own state transitions, not wait for permission. Event-driven > orchestrated
sailhouse.dev/blog/autonom...
#AI #Agents
Lol
Ah sorry - as in investing in CLAUDE.md files et al.
I’ve only made use of it in established codebases as that’s where I’ve seen it be a real force multiplier (utilising existing patterns, components etc.)
Been trialing @zed.dev past 2 weeks in agentic mode - it’s a great integration, a bit eager though, still happy with it.
Claude code is my all round favourite - can truly leave it alone and it gets on as I’d expect.
Requires good investment in documentation though, but well worth it!
Not every event-based system needs a streaming platform.
Kafka’s great if you’ve got a team.
Most of us? We just want something that works, retries, and doesn’t explode in prod. What do you use?
I literally remember a moment 12 years ago when I realised that computers understanding semantics was the breakthrough moment, and here we are.
I did this the other week with a GitHub label, ie the presence of the label triggers a workflow that creates the preview env
I've spent the past 2 evenings building full-stack deploy previews for @sailhouse.dev, from @planetscale.com dev branches -> @fly.io application with a full multi-node deployment.
I am now dead inside, but god is it cool.
This can happen with any AI agent… or human developer!
A BIG difference w AI agents: they will write more code, faster than before, and ship a LOT more code to prod! So ship more bugs as well.
Software engineers (not AI engineers) will build o11y solutions to solve for this!
Screenshot of Logi Options, with a "Logi AI Prompt Builder" popup offering the chance to access ChatGPT directly from where you're working.
Feel like Mark Baum in the big short seeing a prompt builder in my mouse software.
They have overtaken every channel, nowhere is safe
They've discovered slack emojis can be HDR-ified at work and my eyeballs fucking hate it
Huge news - extremely happy for you
“The council is not duty bound to provide street lighting” is not a phrase a public body should use
Pooch & Mutt atm!
If your allegory hadn’t come from eastern Europe or southeastern Asia I’d have been shocked
Dog had a good day at daycare
One of the best roles of my career - incredibly proud to see it turn 10
10 years of innovation.
10 years of community.
10 years of transforming the web.
Today, Netlify turns 10! 🥳
Thank you for being part of this incredible journey.
Watch CEO @biilmann.blog reflect on a decade of building a better web, together:
We built an open-source email confirmation system! It uses event-driven architecture via @sailhouse.dev , runs on @netlify.com Functions, sends emails via @resend.com, & requires zero database setup.
Classic "I'll save time in the long run" developer move, right? 🙃
sailhouse.dev/blog/email-c...
Yeah lol
I was wondering if it's vector/pixel level generation of text or whether it renders it in a non-ai way properly and then composites it
that is insane
I've practiced vibe driven development my whole career and even I tire of it
Warm referrals ("I worked with X, they are amazing, we should see if they would be interested in this job we're recruiting for") seem like they will become a lot more relevant as inbound applications become far, far, far more noisy...
What’s a little auth bypass between friends.
