Hasan Genc

coding @hubspot. Previously @microsoft @trendyol

[Hasan Genc @hasangencx]

@DroghedaUnited Don't forget the Kuymak on the menu.

[Selim Dervişoğlu @SelimDervis3]

Trabzonsporlu olmak… Formayı gördüğün an, daha önce ne kadar samimi olduğunun hiçbir önemi kalmaz. Bilirsin ki büyüğün sana abilik, küçüğün de kardeşlik yapacaktır. Gün gelecek yüzlerimiz değişecek ama biz hep var olacağız. Biz ve bizim gibiler. Dünyanın her yerinde.
📍İrlanda

[Hasan Genc @hasangencx]

@steipete Also searching the web for React's latest version feels like the long way around when curl https://t.co/JJ6kY96zv0 would nail it instantly. Web search tools usually the first choice by models when a simple curl is actually all you need.

[Hasan Genc @hasangencx]

@_hyf0 @evanyou @napi_rs @voidzerodev 🔥

[Hasan Genc @hasangencx]

@Cloudflare AI crawlers have used creators' content without compensation for too long. The web needs a fair model.
@X 's creator revenue sharing is a solid example of valuing content properly.

[Hasan Genc @hasangencx]

@steren Would be nice to see how much did the load test cost.

[Hasan Genc @hasangencx]

@G4brym @aninibread @G4brym loving the AutoRAG updates! Any plans for an API to create AutoRAG instances programmatically instead of manually via the dashboard? Would make automation so much easier for scaling workflows.

[Hasan Genc @hasangencx]

@BraydenWilmoth no extra storage or compute costs is huge! 👏

[Hasan Genc @hasangencx]

@BraydenWilmoth @CloudflareDev I also noticed in the docs that location hints are possible, so we can influence where a DO instance is created (though it’s not available in every Cloudflare data center yet). That’s a nice touch for optimizing latency if needed.

[Hasan Genc @hasangencx]

@BraydenWilmoth @CloudflareDev Thanks for the clarification! With this and the docs, my understanding is that each DO instance is created on the nearest available server to where it’s first invoked. After that, all Workers around the globe route requests to that specific instance using its globally-unique name

[Hasan Genc @hasangencx]

@BraydenWilmoth @CloudflareDev Since Workers on edge servers route the initial request, I’m not sure if the DO’s location impacts this stickiness or how CF handles global distribution to keep all users on the same instance. I couldn’t find a clear answer in the docs and would love to understand this better.

[Hasan Genc @hasangencx]

@BraydenWilmoth @BraydenWilmoth (thanks for the post!) @CloudflareDev I’m curious about sticky WebSockets in this setup: with users from different regions, how does Cloudflare ensure their WebSocket connections remain pinned to the same DO instance for real-time messaging?

[Hasan Genc @hasangencx]

Dancing pyramid with background music
- Crazy that it waits for confetti then speeds up.

Codestral 2501 & Pairagent

[Hasan Genc @hasangencx]

This is the 'open' in OpenAI, accessibility, and humanity at its core. Technology should uplift everyone, especially those who need it most.

[Sam Altman @sama]

@BLCNYY @OpenAI we are working to make it much more efficient and then will offer much higher usage limits

in the meantime, please send me your email address

[Hasan Genc @hasangencx]

@theo All models will require a few iterations for this task
https://t.co/XX5x7bigYn

[Hasan Genc @hasangencx]

I just built the same thing using Gemini 2.0 Flash with Pair Agent - took a few iterations (just like OpenAI's demo surely did)

https://t.co/LShfW6FFlm

[Hasan Genc @hasangencx]

With compare feature

[Hasan Genc @hasangencx]

LLM price listing & cost calculator 🫰

[Hasan Genc @hasangencx]

https://t.co/9KmdjxMbVs

[Hasan Genc @hasangencx]

What makes this impressive:
• Collision detection with rotating boundaries
• Vector math for accurate physics
• Handling edge cases (balls staying contained)
• Translating natural language to mathematical implementation

This is far beyond what AI could do 5 years ago.

[Hasan Genc @hasangencx]

I just built the same thing using Gemini 2.0 Flash with Pair Agent - took a few iterations (just like OpenAI's demo surely did)

https://t.co/LShfW6FFlm

[OpenAI Developers @OpenAIDevs]

GPT-4.5 with @flavioAd’s prompt. Then we asked to make it more creative.

[Hasan Genc @hasangencx]

It's not *exactly* the same animation as OpenAI's demo, but the point is the same.

The goal here is demonstrating how LLMs can handle complex edge cases and follow detailed technical instructions - even for physics-based animations.