Most AI products tie “messages” to a single chat. But what if messages lived with the thing they’re about? Imagine a page, design, query, or dataset, each with its own conversation history between you and the model. Mini-sessions, embedded right where the work happens, instead

Banger post on good system design by Sean Goedecke, Staff Engineer at GitHub. TLDR: good system design is not about clever tricks, it’s about knowing how to use boring, well-tested components in the right place. Some key insights: > Stateful components can get into a bad

at my prev startup we raised 1.2m fought for 4 years searching for pmf didn't find it and shut down we walked away with nothing actually less than nothing if you count all the credit card debt + unpaid salary but that's the thing. if the ships sink the captains go down with

In the next version of Bun, you can show your console logs to your therapist directly from the CLI

Some gems from my weekend reading list: 1. What are OKLCH colors? (jakub.kr/components/okl…) 2. Simulations: the secret behind every great agent (sierra.ai/blog/simulatio…) 3. Hire people who give a shit. (alexw.substack.com/p/hire) 4. Building a web search engine from scratch in

People on 𝕏: “MongoDB flopped. No one uses it anymore.” Reality: - $591M revenue in Q2 FY26 (+24% YoY) - $420M gross profit in Q2 - $2.3B cash on hand (as of July 31, 2025) Wild how often the narrative here diverges from actual numbers. finance.yahoo.com/news/mongodb-i…

"Despite $30–40B in enterprise investment into GenAI, 95% of orgs are getting zero return. Not because of model quality or regulation, but approach." Anecdotally, I agree. We see this every day in our work. The bottleneck isn’t models, it’s system design. Think full-stack

"sometimes you need to study barnacles for 8 years before you can publish evolution" "at the moment when i put my foot on the step, the idea came to me" - Henri Poincaré (French mathematician) "the big AI labs don't fully know the capabilities of their models, advertising

This past weekend's reading list: > Do the simplest thing that could possibly work by Sean Goedecke (seangoedecke.com/the-simplest-t…) > The Muslim Technologist (jihad.house/posts/the-musl…) by Jihad Esmail > Autonomous networks (meter.ai) by Meter > The Coding Agent

Would be nicer if these reminders were sent closer to the date Cursor It's annoying to see them every time I start a chat

so much of building a great business is about believing in yourself and just iterating consistently fuelled by good inputs (i.e. insights, conviction, user feedback, etc.)

Today, we're launching Basedash Agent: your new AI data analyst. It can answer questions, create visualizations, and surface insights on all your company's data. producthunt.com/posts/basedash…

500M lines of code reviewed this month is insane. Congrats to the team.

Waiting for an LLM to say "You're absolutely wrong!"

The onboarding time to a new codebase has significantly shrunk with Cursor. Instead of reverse engineering the code, you can just ask it pointed questions. You can also get started without knowing the full context because you can just search for items as you need them. "Is

Today, we're introducing the Build System 2.0 for E2B. We've completely rebuilt how custom sandboxes are made from the ground up while being focused on 4 main goals: 1️⃣ Better DX 2️⃣ Agent-first experience 3️⃣ Speed 4️⃣ Painless Migration 1 / 🧵

It’s incredibly fulfilling to watch your customers win.

If you're building a collaborative SaaS, implement the org/workspace/project/role/member hierarchy from Day 1. It's often overlooked and is much hard to retroactively add. This hierarchy may have a different name in the frontend, but in the backend, it's always a similar

Every few months I rewatch this guide on sync engines by Tuomas at Linear and every time I walk away with an insight I hadn't picked up before. youtube.com/live/WxK11RsLq…