Darel Bitsy

CTO & Head of Product

[Darel Bitsy @darel_bitsy]

From How to Why: What I kept when the technical lens stopped being enough

https://t.co/DQIDrjOOzV

[Darel Bitsy @darel_bitsy]

@Gemini_Notebook Hey @Gemini_Notebook could we please have duplicate source detection and source count up to 500 ?

[Darel Bitsy @darel_bitsy]

@OpenAIDevs Is this not available in the UK? Codex is not showing it.

[Tim Sneath @timsneath]

For web developers on Mac, good news! Safari is adding an MCP server so your agent can easily test, preview, and debug web content with @webkit. For example, @claude can now see what it's building without needing Playwright or other tools. More here: https://t.co/xQ38tQ2kGO

[Darel Bitsy @darel_bitsy]

@simpsoka Plan mode.

[Darel Bitsy @darel_bitsy]

@thsottiaux Plan mode

[Darel Bitsy @darel_bitsy]

@thsottiaux Plan mode really need improvement.

[DeepSeek @deepseek_ai]

🚀 DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length.

🔹 DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
🔹 DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice.

Try it now at https://t.co/GCdiMzk1Dl via Expert Mode / Instant Mode. API is updated & available today!

📄 Tech Report: https://t.co/drlDrxkYtp
🤗 Open Weights: https://t.co/T13Y8i7SDM

1/n

[Darel Bitsy @darel_bitsy]

@Alibaba_Qwen @huggingface @Alibaba_Qwen 2b and 4b models please, for on-device inferences.

[Darel Bitsy @darel_bitsy]

@Parmarsg @googledevs @GoogleDeepMind Have you check the source code of the gallery app:

https://t.co/a8bL9S1dUY

[Darel Bitsy @darel_bitsy]

@GOROman What’s the format of the model ? LiteRT or ?

[Darel Bitsy @darel_bitsy]

@Prince_Canuma @ivanfioravanti @GoogleDeepMind This is great!!! Thanks Prince. Looking forward seeing it ported to Swift. This is will be key for our on-device voice agent 🙏🏿

[Darel Bitsy @darel_bitsy]

@adrgrondin @ivanfioravanti Would be great to integrate tool calls . In the app. Something like search for example.

[Darel Bitsy @darel_bitsy]

@Prince_Canuma Interesting!! Amazing work!
Let me convert the KV cache optimisation to my fork of MLX Swift so it can help with the work I am doing with inference on mobile device.

[Darel Bitsy @darel_bitsy]

@jtdavies Hi John, any chance you could publish the changes? I have tool calling working reliably on my version but only with fine tuning and that’s without counting the hallucinations. So really interested to see your approach to make the tool calling more reliable for such small model.

[Darel Bitsy @darel_bitsy]

@ChujieZheng Qwen-3.6-2B and 4B for edge devices

[Darel Bitsy @darel_bitsy]

@Prince_Canuma @ivanfioravanti @GoogleDeepMind @TIIuae @IBMResearch @facebook Would be great to merge these into the swift version of mlx

[Darel Bitsy @darel_bitsy]

@Prince_Canuma @ivanfioravanti Is there a swift package for this ?