Hey HN!

My friend and I made and just launched Speq: A collaborative web-based repository for your product's specification. It's a tool that peppers you with questions about your new project until it (and you) truly understand what you are trying to build. Then we put everything together into a comprehensive Speq (see an actual example here: https://getspeq.com/#anatomy) that you can then turn into a PRD, share with colleagues to evaluate and collaborate on, or hand off to an agent to build via MCP.

We both left our Fortune 500 Eng Leadership roles last year in hopes of spending more time with family and getting experience with AI projects. Having spent the majority of our 18+ year careers cranking out feature work at a high-velocity, we wanted to take a stab at fixing the "SDLC". Little did we know that things would move so fast over the next several months that the SDLC as we knew it was on the brink of being totally unrecognizable. We spent a few months contemplating what we could build that would be durable in this new era where the rules are being rewritten, and the big model companies are Sherlocking products left and right.

We stumbled upon GH's SpecKit late last year and fell in love with the concept. It felt like a refreshing approach to an age-old problem. We liked a lot of what SpecKit did, but struggled with the idea of keeping everything in MD and in the repo. Our experience working with product partners and designers led us to believe that a web-based system that allowed for collaboration while still taking advantage of LLM intelligence might be a better solution. So we took the plunge and built what we wished we could have had over the last several years: a collaborative tool that owns and shares the source of truth for your product's requirements to anyone or anything that needs it.

We had no idea what the market conditions would be like when our product was ready to launch, but it actually feels like really good timing. As Tibo (Codex lead) said on X tonight, "the value of good instructions has never been higher" (https://x.com/thsottiaux/status/2050811900062286168?s=20), and that is what Speq is all about.

Speq helps you define exactly what you want to build (defining your vision, navigation flow, product requirements, logic, and tech requirements) and makes it really easy to share with agents and people alike. And it's really easy to edit, version, and evolve the Speq over time as well. When you couple this with bleeding-edge AI models that are insanely great at building things now, you can get more work done, and with better quality, than we ever could have dreamed of.

So please give it a try! Thanks to Microsoft's generous support, the tool is free for the time being. We'd love to hear thoughts and feedback, and to make this a more valuable tool for everyone.

The Lore of Sam Altman Is Being Tested Like Never Before

Wikipedia in the Terminal

The Bureaucratic Escalator and how it operates

Does APL Need a Type System? (2018) [video]

Text Files as a User Interface

Matt Mullenweg thinks WordPress is in decline. He may be right

Valuation Spaces and Relativisation: The Lambda Calculus Example

ChatGPT Wrestles with Its Most Chilling Conversation: How Do I Plan an Attack?

Show HN: Speq – A collaborative web-based repository for your product's spec

Made in China means made in Yiwu

Do LLMs Reason, or Do They Just Predict Math Text?

Investors pile into clean energy as Iran war drives push for energy security

MCPages

Thoth – open-source Local-first AI Assistant

Comparison: Different AI systems structure decisions under real-world scenarios

MPA Renews Push for U.S. Site-Blocking Legislation, Citing Live Sports Piracy

Show HN: Apple's Sharp Running in the Browser via ONNX Runtime Web

Awesome Vintage LLMs

Technology helps observers understand how iconic artworks were created

Packages release more often than ever. Or do they?

Enable password input feedback for sudo

Leeds Town Hall Organ Renewal

Startup request: MCP household appliances

Dear Richard Dawkins

Researchers print structural colour with an inkjet printer

Clawback – rehearse OpenClaw upgrades before touching your live install

Wire-level context pruner for Claude Code

NIST's CAISI Evaluation of DeepSeek V4 Pro finds it to be on par with GPT-5

OpenAI: Auto-review of agent actions without synchronous human oversight

Definable Real Number

Show HN: BoxLite – the all-terrain micro-VM