Over the weekend, while my partner took kiddo out for a play date, I was able to finally get onto a personal project and managed to build the basic backend & frontend in a couple of hours and have both running on localhost. BE: Python+FastAPI, FE: React+TypeScript #buildinpublic

It’s been 2w since I touched my project. What happened: daycare staff out sick, called in Oma backup. Kid got sick, then me. Work overflow, Halloween party, and now my only free time? Feeling crummy from cold symptoms and lack of sleep. Welcome to the glam life of working parents

For a personal project, what tool to use for project management? As everything for my projects lives in private GH repos, I just use GH Projects which connect nicely to GH issues and offer basic agile templates. Zenhub plugin would be a great enhancement but it's not free so🤷‍♀️

"In fact you're usually suffering from three kinds of ignorance simultaneously: you don't know what makes you happy, what the various kinds of work are really like, or how well you could do them." -- Paul Graham

If you’re too lazy to write a prompt, try meta prompting - in OpenAI playground, pick a GPT4 model, click on “Generate”, write 1 sentence to describe the task. Then you get a full length system prompt that can work as the nice first version to iterate on✨ #chatgpt #prompting

You’d think after 50+ years of email (and Apple+Gmail taking up 84% of the market) there’d be nothing left to reinvent. But interesting new email products still pop up. Most recently I noticed Superhuman - a very interesting attempt to make email more productive and collaborative

Just realized today: MCP will be the new entry point for SEO. In the past companies optimized for google search; in the future everyone will optimize for AI search - via remote MCP servers. #mcp #ai #llm

Building something so compelling users choose it over YouTube or ice cream — think about it before bed, wake up at 5am to use it. That’s the dream for any builder. That’s what using Claude Code feels like for me.

If your AI app was built over 1 year ago, it might be time to throw it out and start fresh. MCP, AI agents, reasoning models, claude code—none of them existed then. Don’t patch the legacy. Rethink and rebuild.

Atlassian products are the opposite of AI-first design.

LLMs share many shortcomings as humans, e.g. tendency to overengineer, to fix the symptom instead of root cause, to lose track of longer context. So when you complain about LLMs, just remind yourself that they are trained on human data after all 🤧

1. Take the agent’s point of view. 2. Don’t ask the agent to achieve what it can’t measure. 3. Don't ask the agent to know what it can't verify. — Rich Sutton

we should stop trying to find simple ways to think about the contents of minds. They are not what should be built in; instead we should build in only the meta-methods that can find and capture this arbitrary complexity. The search for them should be by our methods, not by us. —RS