This is incredible news, especially for the AI Agent Engineering & Autoresearch researchers. An overview: SIA (Hebbar et al.): an open-source agent that rewrites both its own harness AND the underlying weights. Results: > 56.6% gain on LawBench > 502% improvement on single-cell RNA denoising > 91.9% runtime reduction on GPU kernels Let's host an emergency hackathon at the house this weekend like we did for @karpathy 's Autoresearch! @kunalbhatia91 !

Never knew this meaning

Any alternates to GitHub?

I think at this point, fewer LOC should have more value. AI creates sloppy lines, you can feel proud with it as more 'work done' but it's a fallacy.

People with lot of green spots on your github contribution chart, do you prefer quality or quantity?

I see so many on device solutions coming up. Like on device LLM? The whole purpose of cloud was to have complex compute done remotely, cheaper and efficiently. What a UNO reverse on business models.

I am participating in the @claudeai Opus 4.7 Hackathon was building stuff related to Voice AI and memory. But given the crazy amount of security attacks @Bitwarden @vercel etc etc I might want to create something for this. Right @getdelve ?

Building cool projects with STS realtime audio model. Any requests?

We can’t. That’s the funny part. So basically the past that has been catalogued can be assumed as open to use. Only thing we can do is make things future proof. :)

The smartest lobster. @thehousesf

Apart from LLMs, people need to start taking a look into security, especially cryptography and PQC. No one is going to boast about they made a quantum machine that broke an existing cryptographic protocol or they have the information stored waiting for an advanced machine that can break your encryption. We need to transition to post quantum cryptography faster.

autoresearch by @karpathy is game changer. I used the idea at Auto Research hackathon @AGIHouseSF for Kaggle competitions to improve models iteratively with agent in loop. The goal is to have agents do testing and trials for hyper-parameter tuning, feature engineering and more, given you have enough resources/GPUs. Hopefully we get some unique solutions to some very difficult problems. Shoutout to @hyperbolic_labs for making this happen.

Just saw @hwchase17 at the Daytona Conference. Legend.

hiring people for ads company. I need you to have software skills and should know how to use AI. not for vibe coders. in-person, in SF we build at high pace with people with high endurance. Dm.

A 0.1 version upgrade in an LLM model should not be hyped this much. Something is wrong. @OpenAI @AnthropicAI

Youtube Video > AI Response

RAG Federated learning for private companies is a great solution!

A* A more well known algorithm for finding distances between two points. Well one of these best path finding algorithm that uses Heuristics to find the most optimized goal and faster. It calculates the cost of travel up to the current point and add the hunch cost till the goal. This way it find a optimized and fast cost to the goal without exploring all the possible paths(and even missing shortest path, unlike Djisktras). A* has few cons like high space complexity, and hence faces issues say in 3D spaces, though there are optimized versions which help offset the high space costs. This algorithm is often used in navigation, for AI in games, robotics and much more. #algorithms

The Voronoi Algorithm: Mapping Proximity One of the most beautiful algorithms and pattern creator. It's in nature -> Giraffe’s spots. Voronoi Diagrams are the gold standard for spatial partitioning. The algorithm divides the space into cells, where every point inside a cell is closer to its own seed than any other. It effectively maps the "territory" of a data point. In Machine Learning, this is the backbone of the 1-Nearest Neighbor classifier. The Voronoi edges become decision boundaries. If a new input falls into a specific cell, the AI classifies it instantly based on that cell's seed. Drones use Voronoi edges to find the "max-clearance" path—the safest route furthest from obstacles.

Minimax Alpha Beta Pruning : Day 1 Minimax is a decision making algorithm designed in 1928 by John von Neumann(great mathematician), helping with game theory to play two player games, like chess or tic-tac-toe etc. It creates game trees, calculating the maximum values and minimum values both players can achieve, which allows us to evaluate all the strategies possible and find the best win. It's application also went beyond the board games to military operations and poker games. There's even a modern LLM named after it: @MiniMax_AI . Alpha Beta pruning is an optimization that prunes the branches that are not 'feasible' hence a great way to optimize large trees. Imagine applying this to chess or go, anyone wants to guess the number of possible combinations?