Kushagar garg

@NVIDIAAI Such an informative session you guys hosted tonight! Looking forward to more of this

@finkd When will it be available in India?

A lot of people asked for this, so Muse Spark 1.1 is now on OpenRouter.

@NVIDIAAI @hardmaru Trust me routism is better than fugu

@hardmaru 🤝

We’re excited to collaborate with NVIDIA to build the next generation of Fugu orchestration models together, by incorporating leading open-weights models.

@Bitflicker64 Mine is also growing💪

"Your graph is cute."~ This is what happens when you actually link your shit instead of hoarding notes like a digital hoarder. If yours doesn’t look this connected, don’t even bother replying.

Yeah well routism is better.

if that sounds useful: github.com/Dreamstick9/Ro… ./install.sh then routism point cursor/hermes/whatever at localhost, crank the timeout, stop babysitting one brain at a time

live northstar, 4 tasks, absolute 0–10. conductor vs best solo worker in the pool (strict wins). won 3/4 (75%). mean 8.39 vs 7.86 (Δ +0.53). lost one hard on an api-pipeline job. small n, real pool, dumps in eval_results/. signal to ship—not a victory lap.

honestly tired of agents that are just “pick one model and pray” you know when the answer is mid and you just… open another tab and try a different one that’s not a workflow that’s coping

Knowledge shouldn't belong to a few.

@NVIDIAAI Can this work with Hermes Agent?

NVIDIA DeepStream 9.1 is here, with 13 agentic skills for building video analytics pipelines. Instead of manually building your vision AI pipeline from scratch, describe what you want in plain natural language. Use skills with a coding agent, like Claude Code or Codex, to easily handle setup, configuration, and execution. New skills include Multi-View 3D Tracking (MV3DT) for tracking objects across multiple cameras, and AutoMagicCalib for automatically calibrating a camera network. This release also brings NVIDIA JetPack 7.2 support for edge deployment on Jetson Orin and Thor. All open source on GitHub, check it out: nvda.ws/4vxQ6Yk

Can't wait to use this!

Billions of parameters. Finite resources.

@Bitflicker64 Just 12% bro😭

Be honest... what % were y'all at before they reset the limits? 😭 I only got to 26%

Such a flex to have this!

This is my "feel the AGI" moment: I used GPT-5.6 Sol to train my own autocorrect model that outperforms GPT-5.6 Sol (wtf??) I have no ML background. I have no idea what I'm doing. I just kept pushing Sol until it spat out a SOTA model. And I spent $0. The motivation: Years of talking to AI have made me terrible at typing. Rather than fix my skill issue, I decided to throw more AI at it. My idea was: instead of autocorrect that interrupts my flow, I want to type fast with mistakes and have AI clean it up after. I wanted the smallest local model possible, for speed, for battery life, for science! So I decided to train my own. Inspired by @karpathy’s autoresearch, I ran Codex /goal with this setup: pick an experiment, try it, record the results to a doc, throw it out if it fails, and plan the next experiment without repeating failures. I gave a few examples that had to pass, tight latency targets, and let it run. Sol did some amazing things. First, it scanned benchmarks and shortlisted base models: Qwen 3.5, Gemma 4, Liquid LFM 2.5. It found a dataset on HuggingFace for typed text. Then it built a simulator for fingers striking a Mac keyboard, modeling the physical layout with a Gaussian distribution around each key. It simulated striking the wrong key, wrong order, fat-fingering, etc. With the models + data + simulator, it fine-tuned using MLX right on my MacBook. It had a working prototype within an hour! But accuracy was pretty poor. — Problem 1: Tokenization Sol read papers, ran tests, and identified that the tokenizer was the bottleneck. Tokenization makes typos hard for the model to see, so it memorizes mappings instead of using its language priors. Sol tried ByT5, Google’s tokenizer-free byte-level LLM. This made a big improvement, but the model is old and lacked the knowledge needed to reach Sol performance. Sol dug deeper and realized a tokenizer-free model isn’t needed; instead, it used T5Gemma, an encoder-decoder model. This can understand the input deeply before producing output, and furthermore, Sol could post-train the encoder to improve performance. This gave a much higher ceiling. — Problem 2: Loss function Now the model was correcting some typos perfectly, but ignoring most. Sol realized that standard cross-entropy loss was teaching the model to avoid edits, because the vast majority of characters in the training data were left unmodified. The fix was wild: Sol wrote a custom loss function that byte-aligns the source and target strings, uses a dynamic programming algorithm to compute the minimum edits between the two, then weights correct edits much higher than copies. After a lot of tuning, this dramatically improved accuracy. — Problem 3: Autoregression One failure mode remained: if the model made a mistake, it couldn’t backtrack. It could only predict the next token. Teaching it to “think” like a reasoning model would solve this, but would be far too slow. Sol found a beautiful solution: instead of greedily predicting the next token, beam search over all possibilities. This parallelizes the exploration instead of one linear chain-of-thought. At the end, choose the path with highest cumulative log probability. This worked great, but made the experience worse, since the user wouldn’t see progress until the whole search was done. To fix this, Sol made a clever observation: after each search step, the longest common prefix among surviving branches is guaranteed to appear in the final result, so it can be displayed immediately. As the search progresses, weaker paths are dropped and the prefix grows, so the user sees continuous progress. Sol built all this as a custom MLX pipeline that does the parallel decoding on the MacBook GPU, with just ~40ms TTFT. It’s crazy fast and entirely local. — Final eval (error reduction rate, higher is better): - Apple autocorrect: 49.66% - GPT-5.6 Luna: 82.47% - GPT-5.6 Terra: 87.64% - GPT-5.6 Sol: 90.56% - Our model (1.7B): 91.02% Final cost: - 1 quota reset (thanks @thsottiaux) - $0 (And yes, I verified there's no cheating. In fact, we test words scrubbed from the training data to prove the model isn’t memorizing) There were a ton more details and tangents I could write about: contrastive learning, GRPO, DPO, dynamic masking, and more. Sol is a fascinating and creative model. It blew my mind so many times. Don’t let a lack of experience stop you: Sol makes AI experiments accessible to anyone!

THANKSSS FOR THE RESETTT @elonmusk !!!!!!

omg thanks for this like!!

OSS for the win!!