Deleted every todo app off my phone, a mini notebook has done more for me than any of them something about physically writing it down makes it real in a way tapping a screen never does.
I spent this week going deep on Qibla calculation. I learned something that surprised me and thought it was worth sharing.
Most prayer apps calculate the direction to Makkah by treating the Earth as a perfect sphere. The maths is clean and it works pretty well. But the Earth isn't actually a perfect sphere.
GPS uses a model called WGS84 which treats the Earth as an oblate spheroid slightly flattened at the poles and wider at the equator. This is the same model your phone uses when it figures out where you are.
So there's a subtle mismatch: your location is calculated using the real shape of the Earth, but the direction to Makkah is often calculated using a simplified shape.
For most people the difference is small we're talking a degree or two depending on where you are in the world. But for someone praying five times a day, it felt worth getting right.
Im now switching to a geodesic calculation that uses the same Earth model as GPS. The difference won't be dramatic for most users but the direction will be as mathematically accurate as we can make it.
Curious if anyone has thought about this before or noticed differences between apps.
https://arxiv.org/html/2512.03271v1
Been working on an internal tool for my SaaS called Tamped this week and got something working that felt worth sharing.
The idea: you @ a bot in Slack with a Linear ticket number. It reads the ticket for context, checks out a clean isolated copy of your repo, spins up a Claude Code agent, runs the task, and posts the result back to Slack.
The full loop:
Slack mention → Events API → webhook server → agent runner → Linear ticket hydration → clean git worktree → Claude SDK execution → result posted back to Slack
What makes it feel different from a basic bot:
- It's repo-aware. It reads project-specific instructions and relevant files so it's not acting like a generic assistant.
- It runs in an isolated git worktree so it doesn't touch your in-progress local work.
- It's bounded. Turn budgets, command allowlists, transcript capture — it doesn't go rogue.
- Linear context is hydrated automatically so the agent understands what the ticket is actually asking for.
Still running locally for now. Next step is moving it toward a proper isolated sandbox for production use.
Happy to answer questions on the architecture if anyone's building something similar.
Hey I’m currently in London for work and need a quiet place to take an important meeting, I usually have access to WeWork on the go but stuck for somewhere this weekend.
Hey everyone, I built Nomad, a curated atlas of working cafés for Lisbon, London, Paris, and Tokyo.
I made it because I got tired of “best café” lists that tell you nothing about whether you can actually work there. Nomad treats cafés more like field notes: Wi‑Fi, outlets, noise, laptop policy, and whether a place is truly laptop-friendly or more of a coworking-style fallback.
It’s completely free, and I’m hoping to expand it if people find it useful.
I’d love feedback on:
- what feels missing
- which cities should come next
- whether the current café info is enough to trust when you’re in a new city
