Pratik Sethi

Senior Software Engineer @Salesforce, Perpetual Learner, All things distributed, DMs open

[Eric Zhang @ekzhang1]

It's really cool that in infra, both traditionally
(1) very arcane, slow-moving technical topics (chip design, ASIC DSLs, RDMA networking) and
(2) very private, expensive and difficult-to-get-right topics (research, RL… like Cognition blog post)

are discussed. No shortage of great stuff online.

Yeah there's a ton of capital expenditure and noise, and it's more important than ever to tell a good story. But everything is getting to be so open :)

[Deedy @deedydas]

Today’s been an absolutely stacked day for AI news. In case you missed it:

1. GPT 5.6 Sol and Sol Ultra dropping tomorrow. Early reviews say it’s not as smart as Fable but very positive.

2. Grok 4.5 launches from Cursor / SpaceX that claims Opus 4.7 quality at 80tps and $2/M in $6/M out price.

3. Bytedance Seedream 5 Pro launches as an image ~#2 and nearly as good as GPT Image 2 at 4x cheaper cost: $0.045-$0.09/image, specializing in edits and infographics. Easily beats Meta’s Muse Image.

4. GPT-Live launches a full duplex non-turn based voice model which allows interactions while you’re talking seamlessly, a huge upgrade in audio AI for consumer

Coming soon:
1. Seedance 2.5 Pro expected to launch soon (early July) to further extend Bytedance’s lead in SOTA video gen models. Will use Seedream 5 Pro as the frame generator.

2. Gemini 3.5 Pro launching soon (July 17). Google seems to have fallen quite behind frontier and people eagerly await to see what Google can put out here amidst a string of high profile departures.

3. GPT-6 rumored to launch soon (Polymarket spikes at August 14) with a new larger retrain, taking aim at Fable.

[Mitchell Hashimoto @mitchellh]

Mind boggling to me that I can make a thing faster and there's always people that ask "but why?" What kind of mentality is that? The pursuit of excellence does not need justification. Also, I find in so many cases, we can't know the impact of an improvement until we do it.

For example, one I've talked about before: Ghostty's high IO throughput has enabled terminal program (emulator and TUI) fuzzing at a speed thats incomparably fast to prior solutions. This has resulted in upstream patches to resolve issues in popular projects like btop, tmux, and more.

Speed enabled that anecdotally example that lifted the tides of adjacent communities that don't rely on Ghostty technology at all. I didn't predict this.

Make things better because they can be better and let the results naturally play out.

[Alex Smola @smolix]

Here's part 1 (of 5) of my short course on efficient LLM inference that I taught at Columbia University. Slides are heavily updated from two weeks ago.

https://t.co/WVCf7mUdkY

[Tianqi Chen @tqchenml]

We taught a brand-new mini-series this year at @SCSatCMU on Modern GPU Programming for ML Systems, as part of the ML Systems course, touching on fun questions like what data layout swizzling is, how to use 3D TMA, and state-of-the-art Blackwell programming. We released a curated online book based on the materials: https://t.co/5ZJg2lySNO check it out

[Mitchell Hashimoto @mitchellh]

https://t.co/hXE5lXLpFT

[NeetCode @neetcode1]

This takes like 2 min to read and is the simplest explanation of the “loops” everyone is talking about.

Why can’t we all just speak in plain English instead of trying to make every single ai coding concept seem bigger than life?

I think we know why but still

[Simon Willison @simonw]

Really looking forward to one of the super-fast custom silicon inference providers like @GroqInc or @cerebras getting GLM 5.2 running

Cerebras has GLM-4.7, Groq is still mostly Llama 3.x and gpt-oss

[cat @_catwu]

On Claude Team and Claude Enterprise, you can now use Claude Code to deploy HTML sites and share these with your teammates!

This has changed how we work internally. Artifacts is great format for communicating architecture changes, data analyses, and new prototypes.

[Deedy @deedydas]

Pretty neat that with one URL change, you can now replicate and iterate on AI papers without having to even provision your own GPUs

[Charles 🎉 Frye @ ICML @charles_irl]

When You were wrapping OpenAI,
I studied 𝕋𝕙𝕖 ℂ𝕌𝔻𝔸

When you were having VC chats,
I mastered 𝔗𝔥𝔢 ℑ𝔫𝔣𝔢𝔯𝔢𝔫𝔠𝔢 𝔖𝔢𝔯𝔳𝔢𝔯

And now that open source AI is here, you have the audacity to come to me for help?

Good! I wrote up some docs.

https://t.co/95XSG31K0P

[Fatih Arslan @fatih]

My latest post on control theory and feedback loops has just been published. I’ll start from scratch and gradually build up feedback loops that are self-healing and resilient, capable of scaling thousands of databases.

Check it out: https://t.co/khsqPD8WmT

[vivek @itsreallyvivek]

https://t.co/ZhOgyq7Vgn

[Charles 🎉 Frye @ ICML @charles_irl]

Last fall, we shared our deep dive on FA4 internals.

But we didn't stop at grokking the kernel.

Since then, we've been developing improvements for inference performance and upstreaming them.

This blog post explains those contributions.

https://t.co/xzDNHdq3Zw

[Cong Lu @cong_ml]

Recursive just came out of stealth, and the team has been cooking 🔥

Our first results: an automated AI research system that can improve AI across 3 very different settings across training and GPU kernel optimization.

https://t.co/2pfwCcffwv

[Silvia Sapora @silviasapora]

After interviewing for Research Scientist roles at DeepMind, Isomorphic, Meta, Cohere and more, I wrote up everything I learned. Technical prep, logistics, negotiation, and emotional breakdowns. Check out my guide: https://t.co/eLh20ggMHW

[Demis Hassabis @demishassabis]

Celebrating the milestone of a massive 150+ million downloads of Gemma 4 with the release of the new Gemma 4 12B model! It's incredibly powerful for such a small model and it’s tiny enough to run locally on a laptop with just 16GB VRAM. Apache 2.0 license - happy building!

[Thariq @trq212]

https://t.co/R6exTuF7P8

[Cursor @cursor_ai]

A great cloud agent experience involves a lot more than moving a local agent to a server.

We've learned that it requires a durable execution platform, a powerful harness, and the tools and infra to give agents realistic development environments.

https://t.co/3xb2kGUjFd

[Charles 🎉 Frye @ ICML @charles_irl]

Writing fast, correct kernels is very hard! For both speed and logic, verification is surprisingly tricky.

Very cool work, and writeup, from @_doubleAI_

https://t.co/vuyVEtpUmN