VP of Ideas

Professional atom smasher. Learning web dev on top of a full time EE job. Afterwork livestream every day(ish) on

[VP of Ideas @VPofIdeas]

Me with code in production at the Large Hadron Collider.

[Adam Rackis @AdamRackis]

Me, with C++ in production in the Spotify Core Layer, and League of Legends client network layer watching you children argue about frontend vs backend

[VP of Ideas @VPofIdeas]

@TheCoppinger I somehow completely missed this arc

[VP of Ideas @VPofIdeas]

Boarding a 12h flight and asking the model to perform an extensive code and architecture review, just to consume the last tokens before the quota reset in four hours.

[VP of Ideas @VPofIdeas]

Procrastinating on weekend plans ended up saving my ass.

[Clash Report @clashreport]

All 50 districts of Thailand’s capital, Bangkok, have been declared disaster areas after heavy rain caused widespread flooding and disruption.

[VP of Ideas @VPofIdeas]

"The Livestream API is available exclusively under the Enterprise plan"

[Developers @XDevelopers]

X Livestream API has been rebuilt from the ground up.

Your entire broadcast lifecycle on X can now be powered by our API: go live, schedule broadcasts, run and moderate chat, and get live events for alerts and overlays. Available now. ���

[Qwen Developers @QwenDevs]

We’ve received so much love for Qwen-Image-2.1 over the past 24 hours, thank you!! Also gotten a lot of questions about the license, especially around model outputs. So here’s the answer:

Outputs are not part of the licensed Materials. Users retain the rights to images and other content they generate using the model.

[VP of Ideas @VPofIdeas]

Number 1 in the world in worst air quality right now 👀 https://t.co/CoY2oMbeDL

[VP of Ideas @VPofIdeas]

@BenjaminHouy How would it work for spaced repetition? You give it a list of words and their previous review schedule and it outputs the most likely word you should review next? (I mean, in theory, since it seems it did not work 😅)

[VP of Ideas @VPofIdeas]

@ivanfioravanti Does it still get stuck in infinite reasoning loops half the time, just faster? The only usable Qwen3.8 I found so far has been the UkisAI one.

[VP of Ideas @VPofIdeas]

@danyay No haze at all! Unlike here in Malaysia 🤣

[Thariq @trq212]

We're adding support for AGENTS.md to Claude Code.

Starting today in version 2.1.277, if there is no CLAUDE.md in a folder, Claude will check for and use AGENTS.md.

You can toggle this behavior in /config.

[VP of Ideas @VPofIdeas]

@BenjaminHouy 😲 GPT Live ? Or Gemini?

[Vaibhav (VB) Srivastav @reach_vb]

what’s the best <3B open weights LLM these days?

something that can run easily in a browser/ extension

[VP of Ideas @VPofIdeas]

HN in panic mode at the OpenAI/Claude/Xai outage while I'm chugging along with my local setup without a care in the world, *chef kiss*

[Aporia @aporia9n]

AI will probably create 100x more software, but far fewer viable software companies.

A personal finance dashboard, trip planner, file converter or one-off reporting tool no longer needs to become a product. You describe what you need, use it for a week, then throw it away.

Software is becoming more like content: infinite, personalised and disposable. A lot of it will be generated, not bought.

What still deserves to become a company will sit on top of something persistent: data, coordination, trust, maintenance or accountability.

There has never been a better time to build software or a worse time to assume yours deserves to become a business.

[VP of Ideas @VPofIdeas]

@philkellr Ask your AI agent to ssh in and fix it for you ��

[VP of Ideas @VPofIdeas]

@BenjaminHouy Ben noooooo! Why are you going back?!?

[VP of Ideas @VPofIdeas]

What could go wrong

[VP of Ideas @VPofIdeas]

@petergyang @JPoehnelt Which doesn't even work. It was much faster for me to ask Qwen to write a few lines of python to interact with Google calendar than to get their vibecoded piece of rust to do the same.

[David @dayonefoundry]

I was listening to a panel yesterday and the guy was talking about how planning is hard and implementation is easy where you can turn your brain off and do the grunt work and it’s like a break. But now we got rid of the grunt work and all our time working is deep thinking so we can’t keep up with working at the same pace. Was really insightful I never thought about it like that.