Aditya Pethkar

[Aditya Pethkar @PethkarAditya]

@mehuljindal18 @aurorax_ai @_shyam_king_ This is gonna be so useful
Crazy!

[Mehul Nath Jindal @mehuljindal18]

Introducing Aurora - the Search Engine for IRL

Have you ever had to go shop-to-shop, or ask your friends and family, or spent hours browsing online for weeks, to find out - which business has what you want?

Could be anything from a renting an Apartment, to a Doctor's appointment? Maybe just your favorite shoe in your size at a store nearby?

Now - you can just Ask @aurorax_ai , and it will tell you who has it!

[Rory Tyer @RoryTyer]

we are living in quite interesting and fun times

[maks @maks_p]

@theo UX -> hierarchies, progressive disclosure, navigation, labels, etc

[Theo - t3.gg @theo]

What are some coding tasks that AI is still really bad at?

[Claude @claudeai]

Claude Dashboards and Claude Motion are in beta today.

Ask Claude to turn your data into live dashboards and your ideas into animated explainers.

[wono @wono_strategy]

Tech companies trying to replace people with AI, only for people to start replacing tech companies with AI has gotta be the funniest plot twist of the century

[Gems of BJP @Gems_of_BJP]

Guys please stop. No matter how much I’m against BJP and this govt, I won’t tolerate a foreign nationalist spreading misinformation about my country. Say it like it is. Ambani is not the Prime Minister of India,

Adani is.

[Elon Musk @elonmusk]

Also, Starlink has proven to be essential for saving lives during natural disasters throughout the world, when all other communications systems have failed, so you would also be helping save men, women and children throughout India. 🇮🇳

[Elon Musk @elonmusk]

Dear Prime Minister Ambani,

Please accept my humble apologies for not realizing that you are the real boss of India.

Naturally, you would prefer to maintain your monopolistic exploitation of the great people of India, but would you nonetheless consider allowing Starlink to compete?

There are many areas of India, as there are in all countries, with no Internet access, which denies children the opportunity for education and for small businesses the opportunity to sell their goods to a global market. Starlink would change their lives for the better.

Thank you 🙏

[Elon Musk @elonmusk]

Starlink is licensed in over 165 countries and has spent five years complying with every single law and requirement of the government of India, so why still no license?

Is Ambani the real boss of India?

[Claude @claudeai]

Introducing Claude Haiku 5.5: the cheapest, fastest, and most capable small model we’ve ever released.

On average, it costs around 75% less to run than Claude Haiku 4.5.

[Bryan Johnson @bryan_johnson]

Creatine didn't build muscle or burn fat in this study. It was water retention.

Exercise wasn't randomized. Participants chose whether to train.

DEXA scans count water as "lean mass". Creatine hydrated muscle cells. It didn't synthesize new contractile fibers.

The drop in body fat percentage was due to the math of adding water weight.

[Guitar Cat + LLM 「Will @ SGLang Summit in SF」 @GenAI_is_real]

Ramanujan wasn’t skipping proofs. He was hiding his chain of thought to prevent distillation.

Hardy: “Show your work.”
Ramanujan: “Sorry, reasoning tokens aren’t exposed via the API.”

[Theo - t3.gg @theo]

Another day of Unemployed Kyle (aka WDS) dunking on me because he has no real career experience.

His main complaint here is that I show multiple instances where I didn't know the depth of how part of a codebase worked.

Sounds bad until you realize it's a codebase with over 300 contributors and millions of lines of code. Doesn't matter how much AI is used, nobody knows the implementation details of all the code in a real world codebase like this.

The example he highlights is one where I was unfamiliar with details of how tagged skills are attached when we pass prompts to the Claude SDK. Funny enough, I know a lot (probably too much) about how we tag skills and how messages are passed over SDK, but this one intricacy was new to me.

Knowledge gaps like this are the norm in real world work. Learning to operate with partial understandings of systems is key to long term success as an engineer, regardless of AI usage.

I'm genuinely scared that inexperienced devs like Kyle are going to trick well-meaning juniors into holding back their careers.

[OpenAI @OpenAI]

Please welcome GPT-6 Sol and GPT-6 Luna to the GPT-6 universe.

GPT-6 Sol and Luna build on the advances behind GPT-6 Astra, bringing much of its strengths into faster and more affordable models to support work at scale.

We’ve also made caching and inference more efficient, and we’re passing the savings directly to you: 50% lower API prices for Sol and Luna compared with GPT‑5.6 promotional pricing.

[Theo - t3.gg @theo]

This is a terrible compaction strategy that fundamentally doesn't understand how compaction and context management work.

Seems like a lot of people are confused so let's break this down.

1. Compaction isn't a filter
The role of compaction is to clean up history to keep the agent focused, not just deleting noise. It should be used sparingly when context gets too long, not constantly to keep context small.

2. Jev doesn't even know what it's deciding on!
Models use the context of the thread to decide what to keep or not keep in a summary. This implementation goes through on a "line-by-line" (per tool call) basis to decide what should be left or deleted.

Not only does this 32k token context model know very little of what happened before, but (in this implementation) it doesn't even know what the result of the tool call is!

Deleting these things randomly will keep the model from knowing what it's tried and dooms you to end up in "stupid loops" where the model keeps trying the same thing over and over.

3. You're giving up the reasoning entirely
Frontier models from OpenAI, Anthropic, XAI, and Google do not share reasoning traces over the API. They share encrypted payloads, which Jev cannot see (and often will drop).

Anthropic is even stricter with this, requiring you to preserve the entire history in order to get any of the reasoning data. As a result, using this in Claude Code guarantees the model will act way dumber.

4. Models are tuned on their compaction flows
For the last year, Frontier Labs have been including compaction and long runs as part of the training process. These models have learned ways to compact that are more effective than any rudimentary solution.

Fun fact: If you switch models in Codex and compaction is necessary, compaction will run on the model that was previously used in the thread.

5. Cache writes are more expensive than cache reads.
Cache writes are the biggest cost by far for agents. I often see cache write costs go over 60% of my total LLM spend in my personal use of Claude Code and Codex.

Cache writes are insanely expensive when data earlier in the history is changed (because the old cache is invalidated when things change at the top). Every history edit requires a cache rewrite for ANY data past the history edit.

If your history is "1,2,3,4,5,6" and you delete "2", you have to rewrite "3,4,5,6". This is more expensive than leaving "2" in the history.

Good news. Since we're already killing all of the reasoning tokens by doing this stupid compaction strategy, the rewrite cost won't actually be that high because the model is missing so much data! 🙃🙃

6. The implementation is hot garbage.
> "Whatever is not kept is deleted permanently, but the assistant can always re-run a tool or re-read a file."

Good luck with that one.

To be clear: this is a cool experiment and I find it genuinely interesting. That said, if you think this style of bs filtering on a probability threshold is actually a compaction strategy, I highly recommend you just use the defaults in tools like Claude Code and Codex.

You're much less likely to hurt yourself that way.

[OpenAI @OpenAI]

We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.

The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.

The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.

[Theo - t3.gg @theo]

Hey it’s me, Anthropic’s #1 hater reporting in

Opus 5 is a new model trained on what they learned from Mythos. Hell, it was probably largely disrupted by Mythos.

Let’s say, hypothetically, Mythos is a big model and as a result of training, it has 100 “capabilities”. Most of them are good, but 5 of them can be bad. This happened because they weren’t working from a fixed set of capability, they were trying to squeeze in as much as possible.

Opus 5 starts from Mythos. They looked at all “100 capabilities”, said “okay we only want these 95 of them”, and then trained hard on those 95 through distillation.

Opus 5 is a much smaller model than Fable/Mythos. They effectively trimmed out the “scary” parts. The alignment benchmarks confirm they were pretty successful with that.

I see how it could look contradictory, but I don’t think it makes sense when you reason about it a bit