Dean Sacoransky

Head of FDE @tavilyai

[Dean Sacoransky @deansacoransky]

tavily-research is SOTA on DeepResearch Bench 🏆 outperforming Google, Salesforce, OpenAI, Anthropic, Kimi, Alibaba, Nvidia, Perplexity and others.

We shared our insights in a blogpost - live on @huggingface , read more below👇

[Dean Sacoransky @deansacoransky]

@lakshyaag this approach isn't bitter-lesson pilled

[Dean Sacoransky @deansacoransky]

i would like to see a super high context limit model

[Dean Sacoransky @deansacoransky]

some people are building at the model layer, others applying ai at the application layer. now is time for the organizational layer

[Dean Sacoransky @deansacoransky]

there is no better enabler for ai governance, traceability, and human in the loop than git

[Dean Sacoransky @deansacoransky]

@AnthropicAI @bcherny @_sholtodouglas here is my wish list, and it does not include Fable

- realtime audio on claude code mobile
- 10M context window model
- increased daily limit for scheduled routines
- hide intermediate tool calling and thinking outputs

[Dean Sacoransky @deansacoransky]

tokenmaxxing is the only way forward

[Dean Sacoransky @deansacoransky]

Crazy times and difficult to imagine what one year out looks like, let alone 5-10

[Anthropic @AnthropicAI]

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.

The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.

Access to all other Claude models is not affected.

We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.

Read our full statement: https://t.co/bwn0sximKZ

[Nathan Lambert @natolambert]

If anthropic can't convince a bunch of tech bro's on X that they're not safety washing, good luck convincing the american public.

[Dean Sacoransky @deansacoransky]

@Vtrivedy10 The most underrated “traces” are your personal historical sessions in CC/codex/whatever you personally use. Production agent traces are great for your company’s IP, but personal traces are great for you as an individual

[Dean Sacoransky @deansacoransky]

@t_blom The solution is simple: recording devices and buy in

[Dean Sacoransky @deansacoransky]

@levie Knowledge and code must live together...and a code execution environment is required. This applies to both technical and knowledge work

[Dean Sacoransky @deansacoransky]

@dbreunig Great post and really resonates! Human docs should be more focussed on product and machine docs on engineering. Not sure the best way to convey a mental model is through markdown tho...

[Dean Sacoransky @deansacoransky]

@willccbb @martin_casado closing the gap in-domain is not cheap for the average developer or business person…if it was, people would be fine-tuning models more often

[Dean Sacoransky @deansacoransky]

100% locked in on building a central, company-wide environment, code and knowledge living together. The technical problem is super simple. The real challenge is a human problem and it’s way harder to solve: buy in, change management, bloat.

[Garry Tan @garrytan]

This is the actual bottleneck. The models are smart enough already. What is missing is the company-specific context locked in senior people heads. Whoever cracks knowledge extraction at the company level unlocks the rest.

As you work on this, please consider using GBrain as your OSS retrieval layer

https://t.co/0F5uDQzPHu

[Dean Sacoransky @deansacoransky]

@garrytan There is no need for an extra retrieval layer. The models are capable of navigating the company-wide knowledge base

[Dean Sacoransky @deansacoransky]

@Vtrivedy10 tokenmaxing is the future

[Dean Sacoransky @deansacoransky]

@SergioPaniego @huggingface What’s the term for orchestrating many coding agents in parallel?

[Dean Sacoransky @deansacoransky]

@dbreunig What exactly do you mean by “Memory systems”

[Dean Sacoransky @deansacoransky]

@nofil_ai if you’re not using jetbrains your ngmi