srujan

[srujan @srujxx]

@mattpocockuk Need for codex

[srujan @srujxx]

@milesdeutscher @rebryk

[Eric Berger @SciGuySpace]

This seems like a pretty big deal.

https://t.co/31Ym1UJvK3

[OpenAI @OpenAI]

Now in private beta: Aardvark, an agent that finds and fixes security bugs using GPT-5.

https://t.co/xwtJhfDM3X

[srujan @srujxx]

@shrix_x @100xengineers yep, atb

[srujan @srujxx]

@ddcaridi thank you, you too on the 2m

[srujan @srujxx]

@supermemory Persistent memory for long-term for a companion?

[srujan @srujxx]

@supermemory is it diff is it from letta?

[Letta @Letta_AI]

🧱Memory Blocks are an abstraction for managing an LLM agent's context window. This concept, originating from the MemGPT research paper, provides a systematic approach to organizing what information an agent has access to during inference.

[srujan @srujxx]

🔒

[Yurii Rebryk @rebryk]

Pre-seed is tough. No users, no revenue, just an idea and a deck.

These 18 startup programs can help founders to build from zero:

1.) @spc , US
2.)@ycombinator, US
3.) @HF0Residency, US
4.) @seedcamp , Europe
5.) Arc by @sequoia, US / Europe 🧵

[conor brennan-burke @contextconor]

ADHD CEO + autistic CTO is the most cracked cofounder combo

[srujan @srujxx]

@shweta_ai It's not like they bypass ads, the ads promoted on top get the agent's first impression

So Google is not cooked but just losing market share,

However, they are cooked cause of Gemini being Gemini

[srujan @srujxx]

@sama I don't know about this one

[srujan @srujxx]

🔒

[Ben Lang @benln]

For any pre-seed founders exploring accelerators, this is my shortlist:

• @HF0Residency (SF)
• @spc (SF/NYC/Bangalore)
• Arc by @Seqoia (US/EU)
• PearX by @pearvc (SF)
• @AIgrant (global)
• @Neo (SF)
• @conviction Embed (SF)
• @ycombinator (SF)
• Product Market fit method by @FirstRound (SF)
• Velocity by @VillageGlobal (global)
• Speedrun by @a16z (SF)

[srujan @srujxx]

@youkhna But is it a AI phone though?

We need a Minimal AI phone, like a perplexity assistant in it

[srujan @srujxx]

@AravSrinivas

[srujan @srujxx]

@benln @HF0Residency @spc @seqoia @pearvc @aigrant @neo @conviction @ycombinator thank you for this, you rock

[srujan @srujxx]

@MikeEndale @cursor_ai @windsurf_ai Rollout still active?

[srujan @srujxx]

@AskPerplexity Can you go into details? Special tokens, vocabulary size, etc etc

[srujan @srujxx]

@AskPerplexity Token and embedding in dLLMs @AskPerplexity

[srujan @srujxx]

@AskPerplexity How does token steam differ in autoregressive models vs diffusion language models? @AskPerplexity