Lukasz Jagiello

This is a fun one! One of our engineers uses Friday to manage her D&D campaign. It tracks characters, storylines, and world state, and keeps everything consistent over time. Turns out Friday makes a great D&D assistant. medium.com/p/1fb9dd7ecd7e

Most AI tools forget how you work. Friday doesn't, with Skills. Add your preferences once, and they’re applied across everything you do. Get more consistency, with less repetition. medium.com/friday-ai/01...

Built some AI skills for CTFs → tested at #BSidesSF CTF. 🤖 Agent solved 50/54 challenges almost effortlessly. Feels like a real shift in security problem-solving. github.com/ljagiello/ct...

Skills are one of those things that have dramatically changed how we use AI tools. The difference is night and day and without them, it would be hard to build efficiently modern software. Check out some of our team’s favorite skills! blog.hellofriday.ai/our-teams-fa...

People are trying to force AI to be deterministic. That’s a mistake. It may work at first, but it breaks over time. LLMs shouldn’t generate systems, they should make decisions about them. Let models plan and deterministic systems execute.

We built a way to track every open deal in HubSpot. It monitors open deals analyzes what’s happening suggests what to do next and follows through. this is what it looks like ↓

Building reliable AI agent systems is less about prompting and more about coordination. Once execution starts, reliability depends on managing state, tools, and failure. Read more 👇 blog.hellofriday.ai/building-ai-...

Most AI is great at responding, but bad at doing real work. Work isn’t one prompt. So how do you make nondeterministic AI reliable? - Start with planning - Structure intent into workflows Result: predictable AI that still feels powerful. More here: blog.hellofriday.ai/what-does-it...

A "simple idea" now generates 100+ pull requests every week fixing small issues across our codebase. That’s 2–3 days of manual work saved weekly. Automation handles the tiny fixes so engineers can focus on the hard problems.

Less than a year ago, using AI to get real work done meant specialized skills, tons of code, and endless copy-pasting between tools. That's why we built Friday - a platform where AI agents collaborate to solve real problems, not just answer prompts. Try it: hellofriday.ai

Tiny gem from Linear founder @artman.bsky.social on the @pragmaticengineer.com "We are building on top of Google Cloud. When you are choosing your cloud provider, you really have two straightforward options: AWS or GCP. On paper, they look pretty similar." [1/5]