Tim Ang

pmm @cohere | prev @PalantirTech @Cambridge_Uni @UniofOxford

🖥️ llms | search | multimodal
🧬 multiomics | ngs | glms
🌿 evo-devo | conservation

[nature @Nature]

A powerful gene-editing technique has been used to edit DNA directly in a person’s body — the first known instance of such a treatment.
https://t.co/hoSR2fXR2V

[Tim Ang @ang_seng_chai]

Transcribe again.

[superwhisper @superwhisper]

Superwhisper 3.0.2 brings faster on-device transcription with @cohere's Transcribe ⚡️

[Tim Ang @ang_seng_chai]

I'm oddly into our Rerank models. Such a simple value add: refine the content you feed your LLM so it consumes fewer and more relevant tokens.

Context Focus is also a great addition. Don't just return docs or chunks - return the exact spans themselves.

Try it out. It's one of the simplest AI cost levers your company can pull on.

[Cohere @cohere]

Introducing Rerank 5

A few lines of code can change how you search. With this release, we’re advancing the frontier in AI retrieval and precision.

Available in Pro and Fast tiers today.

[Cohere @cohere]

Introducing Rerank 5

A few lines of code can change how you search. With this release, we’re advancing the frontier in AI retrieval and precision.

Available in Pro and Fast tiers today.

[Tim Ang @ang_seng_chai]

Octave (formerly @HexagonSIG) uses Cohere Embed on @awscloud Bedrock to power multilingual and multimodal search across their regulated enterprise data

One of hundreds doing the same. Embed offers premium accuracy, lower vector storage costs, and is compatible with your data.

The post: https://t.co/ncIKQ4lwKO

[Tim Ang @ang_seng_chai]

@SebJohnsonUK Everything beats Whisper on accuracy.

[Tim Ang @ang_seng_chai]

@khudonogov @NatWestGroup @cohere Yes - it will boil down to the data. That being said, who knows what our peers have used to train their models.

[Tim Ang @ang_seng_chai]

Cool independent research from @NatWestGroup reporting @cohere Transcribe's performance on British-accented speech.

15.3% word error rate, outperforming Whisper Turbo and Parakeet, with one of the smallest accuracy gaps across accents. All without fine-tuning.

Better speech recognition = more reliable banking voice assistants, customer service automation and voice-enabled financial agents.

Paper: https://t.co/fnSvyITr97

[Tim Ang @ang_seng_chai]

So bleak. Just as Whitehall insiders complain Miliband has been obstructing AI investment due to net zero incompatibility.

I doubt this is a simple question of incompetence or bureaucratic bloat, either. So many from the Labour and Green ranks will celebrate this.

[Bloomberg @business]

Microsoft has pulled out of a UK data center deal with British neocloud Nscale at a critical time for the startup’s initial public offering hopes, the Telegraph reported. https://t.co/yiyXv1PWmU

[Cohere @cohere]

If your company just tightened the token budget, they should probably try out North Admin.

Control your models by the who, what, when, and where, and track and manage tokens in real time.

[isadora @gunpowderandran]

minha parte favorita da bíblia é quando Jesus fala para os discípulos que eles tem que ficar a temporada toda no Real Madrid

[Tim Ang @ang_seng_chai]

@lemuel787 @cohere @Cohere_Labs @bpjzy ?

[Tim Ang @ang_seng_chai]

Powered by Cohere Transcribe and Tiny Aya, by the way. @cohere @Cohere_Labs

[superwhisper @superwhisper]

Introducing Superwhisper Meetings

The first fully on device meeting recorder and note taker.

[Tim Ang @ang_seng_chai]

@shahab_cha @UnslothAI Thanks for sharing that mate. I'll definitely give that a spin. What other local models do you build with / compare against?

[Tim Ang @ang_seng_chai]

Cohere Transcribe is now supported in @UnslothAI.

- Run speech-to-text entirely locally, across 14 languages
- Fine-tune with ~40% less VRAM
- Build audio AI apps with accuracy that reliably beats Whisper and Parakeet

Love Transcribe.

[Daniel D'souza @mrdanieldsouza]

🚨🫠 Throwing more agents at a problem is NOT a scaling strategy 🫠🚨

Introducing Scaling Inference-Time Compute with Teams of Coding Agents from @CohereLabs where we investigate when to delegate work across agents to boost performance without blowing up cost!

More here 👀🧵👇

[Nick Frosst @nickfrosst]

turns out you can get frontier performance by having a swarm of tiny models working on the problem together.

great work from the research team on this one :)

https://t.co/OoTmFrKo7k

[JT @jiratickets]

grown men with computer science degrees after typing "plz fix" and hitting enter

[Heartbreak Feels Good @InAPlaceLikeAMC]

@johnternus The way Nolan intended us to watch the Odyssey. Thank you @Apple !