The personal agents - like Muse, Instinct etc are a bigger threat to large consumer companies than the underlying AI itself. Rent seeking will be impacted in the short term. Good post from @chamath

So, @OpenAI is doubling down on Consumer AI in India.

If you are a founder or planning to become one -this is worth your read.

thanks to everyone who joined my session at @aiDotEngineer paris yesterday! for everyone who missed it, here's the thread form of my talk 🧵

This is such a wonderful watch. At this point, we all are using coding agents to build and pretty much most of us aren't reading the code. How to ship reliable code? A good summary from Matt here - but I would still urge you watch the video (and yes- watch it at 2x speed :) ) nitter.net/mattpocockuk/status/21…

My most productive activity today was cleaning up space in my primary work laptop, a 2020 MBA. I had 12 GB SSD space left. Every app became terribly slow. I told ChatGPT to find ways to free up 80 GB. 15 min later, I have reclaimed 100+ GB. For some of those fixes- am surprised Apple does not have a product to do those!

Many years ago, when I was still employed at a large Corporate, the Engineering director asked everyone working on the project to read the book "Phoenix Project" . This included my AI team too. He even bought some books for the team. That book itself is derived from the classic business book "Goal". The book is all about analogy. Set in a fictional automobile manufacturing plant, it introduces challenges there with what someone in IT would face- especially with focus on DevOps. With so many terms floating around in AI, it might be useful to generate domain-specific books that relate to stuff that happens in their industry (guess we are way past writing these days!).

Great team and great problem statements to work on. Link to apply in the embedded tweet.

Opus 5.5 is fast!

I wish they'd taught Geography like this in school!

Laya is open weights and is quite performant, fast and is cheap. Cool work by @Nandakishorm1 laya.convaiinnovations.com/

My evening music the past 3-4 days have been the amazing fusion music from the Bengaluru Ganesha Utsav. They stream the event Live on YouTube and the quality is quite good!

A lot of health systems have requirements that none of their documents be sent to third-party apps/APIs. While its awesome to have Jev do amazing PII masking engine, its unusable in the healthcare industry for this use case.

“Jev can chat too”

Some really cool use cases for System 1 classification models. I am bullish on usecases involving Browser Use/Computer Use

Even without this, Jev has been quite economical. What's stopping you from building something out using Jev?

no chatgpt reset dropped this weekend eh? sigh...

Introducing Bespoke Nimble: an open data, open model, open recipe for an open Jev. Code and info: github.com/bespokelabsai/nim… Model: huggingface.co/bespokelabs/B… Data: * A new data curation recipe called contrastive data curation. * Slightly change facts to generate negative data. This pushes the model to discriminate better and become a better decision maker. The calibration is implicit. * Didn't do ablations but I think this is a critical piece! * This also means training data doesn't need probabilities. * Data covered 10 categories, and is fully synthetic. * This data is split into train and eval. Training * LoRA finetune of Qwen3.5-9B. * Distillation-free: we use Jev to only evaluate. * No RL yet! Serving * Parallel constrained decoding as suggested by @NielsRogge and @harshagundal. Results: * The post-trained Qwen (Nimble) became substantially better on our curated eval: 66% for Qwen to 90% for Nimble. Jev is at 93%. * 100ms on H100 and free to use on your macbook! Feel the AGI for free. * 2 days of building in public. :) Big caveat is that there is no standard benchmark to measure performance, and it's possible Nimble is much worse on other benchmarks compared to Jev. But it should be better than Qwen! We thank @typesafeai for making Jev and the inspiring discussions in the community. Hope this release lifts all the boats and encourages more research and activity in this space.