I spent the last two to three months looking for work, and partway through I started measuring what employers actually ask for in AI agent roles. The search totals lied. The job descriptions told a calmer, more useful story.
Read post →18 posts tagged with "ai"
I spent the last two to three months looking for work, and partway through I started measuring what employers actually ask for in AI agent roles. The search totals lied. The job descriptions told a calmer, more useful story.
Read post →The app is only one piece of what a client needs now. The rest is the package around it: the repo as context hub, a shared room with an agent in it, evals from the start, CI/CD measured against the DORA metrics, and feature flags - and why a solo builder can offer the whole thing.
Read post →
Giving AI its own workspace makes browser-heavy tasks easier to hand off. From recovering $1,600 in solar payments to turning a research workflow into an API, here is where agents in a box are useful.
Read post →
I used to build my back ends in n8n - APIs, background jobs, queue work, AI integration, all of it. Something happened in January 2026 and now I don't reach for it, or for no-code platforms, at all. Here's the walls I hit and what I build with instead.
Read post →
Taking the agent from the last video and plugging it into a Slack room so it listens to everything, remembers it, and wakes up a real Linux box when it needs to. Supabase for the route and the memory, Railway for the box, Zapier's SDK for the CRM.
Read post →
Everyone's talking about agents. Here's what actually changed: instead of writing a pile of tools for a long-running process, we now hand the agent a whole computer and let it use Bash and the old Unix tools to get the job done. Two real examples, and a lot less magic than you'd think.
Read post →
Your vibe-coded app calls the Zapier SDK, which calls a real API. Run that in a test and it's not a test anymore: it needs a key, it costs a call, and it goes red when the wifi hiccups. Here's how to test it without ever hitting the real thing, and the one reminder to give your AI so it always does.
Read post →
You vibe-code a feature, it works, and a week later it's mysteriously broken again. And every time you want to go live it's the same tedious steps: run the database change by hand, copy the keys, click deploy, hope. CI/CD hands all of that to automation that refuses to ship anything broken. Here's how, using a real repo.
Read post →
I drag a feature card in my app and AI writes the code, runs the tests through GitHub Actions, and opens a pull request. I review, merge, and it deploys itself. Here's how the pipeline is wired, from a dragged card to a deploy, and what to do when a step needs a fix.
Read post →
There are plenty of hosted ways to give an AI a browser. This week I built my own — a little Chromium-on-a-server that stays logged in, spits out clean Markdown, and that any of my AI tools can drive. The real unlock: wrap it as an MCP and every agent I make gets a browser for free.
Read post →
Write one plain-English spec file, hand Claude Code your data, and let promptfoo prove your prompt works — including a trick for running the AI judge free on your own laptop.
Read post →
Tools, MCP, and skills sound like they're competing. They're not. Here's the simple mental model I use to know which one to reach for — and when to just combine all three.
Read post →
It still amazes me that we can vibe code our own apps. Here's how to do it with confidence — keeping it secure, getting it hosted, and not getting scared off by GitHub — all built around the humble to-do list.
Read post →
In 2023 I built basically every AI platform feature you've watched ship since — document collections, multi-model workflows, deployable bots. Here's why being early was the same as being wrong, and what I'm building now that the labs structurally can't eat.
Read post →
People don't only Google anymore — they ask ChatGPT, Perplexity, and Claude. There's a small layer of plain files that makes your site easy for those AI engines to find, trust, and quote you with a link back. I added it to this blog in an afternoon, and the AI did most of the work.
Read post →
Cloudflare's always been an option for this kind of thing. It just used to feel too technical to bother with. A spam problem on another project nudged me to finally try it — and with AI doing the heavy lifting, I wrapped my chat around OpenAI with real guardrails in an afternoon.
Read post →
Automation works — I've watched it save companies hundreds of thousands of dollars. But the wins came from sitting next to the person doing the work, not from dropping in a magic agent. Two real stories on what actually makes it stick.
Read post →
Anyone can drop a chatbot on their site in ten minutes. Here's how to do it without waking up to a $10,000 bill or letting somebody else use your AI on their site.
Read post →