hanncx @hanncx

I believe that the most common enterprise AI applications today aren’t the ones that solve the most important problems or make the most money. The most common applications are the ones that are easiest to evaluate. 1. Recommender system: evaluated by increase in engagement or…

model = learn(data) Synthetic data is great, but it’s not data. It’s an intermediate quantity created by learn(). Data is created by people and has privacy and copyright considerations. Synthetic “data” does not - it’s internal to learn().

These are awesome, keep the ideas coming! Summary of responses so far... Agent Integration and Workflow Enhancement: - Agent Interaction: Siva Surendira (@siva_1gc) suggests automated meta prompt generation and self-reflection capabilities for improving agent behavior. -…

In AI research there is tremendous value in intuitions on what makes things work. In fact, this skill is what makes “yolo runs” successful, and can accelerate your team tremendously. However, there’s no track record on how good someone’s intuition is. A fun way to do this is…

AI agents have the ability to fundamentally change the business model of enterprise software. Today, when you build a SaaS product, the primary business model is to sell seats that are tied to the end-users of your service. For "same store" sales, you can essentially grow at the…

We have a new preprint out - your language model is not a reward, it’s a Q function! 1. The likelihood of the preferred answer must go down - it’s a policy divergence 2. MCTS guided decoding on language is equivalent to likelihood search on DPO 3. DPO learns credit assignment

Handling long context in LLMs is expensive, but can we cut the cost by learning them offline for a specific set/genre of documents? Introducing LLoCO, our new technique that learns documents offline through context compression and in-domain finetuning using LoRA, which archives…

Result 10/11/12: surprisingly, when pre-training good data (e.g., Wiki) together with "junks" (e.g., Common Crawl), LLM's capacity on good data may decrease by 20x times! A simple fix: add domain tokens to your data; LLMs can auto-detect domains rich in knowledge and prioritize.

PiSSA: Principal Singular Values and Singular Vectors Adaptation of Large Language Models Significantly improved finetuned perf by simply changing the initialization of LoRA's AB matrix from Gaussian/zero to principal components of W repo: github.com/GraphPKU/PiSSA abs:…

Function calling won’t scale, at least with the current api interface. Here are my overly simple logic: (1) if you have thousands, or even tens of functions to choose from, the current api would just be too daunting to call (imagine what that function candidate field will look…

swyx @swyx

4 weeks ago

Function calling won’t scale, at least with the current api interface. Here are my overly simple logic: (1) if you have thousands, or even tens of functions to choose from, the current api would just be too daunting to call (imagine what that function candidate field will look…

6 2 48 58K 43

Quick rant about RDBMS, multi-tenancy, and AI. Data warehouses are great if you control the shape of your data and hand-code schema definitions. And they are basically useless for multi-tenant products (like @datadoghq, @splunk, @braintrustdata, etc.) where you have to ingest…

Problems I'd do if I'm to do a startup again (though I probably won't any time soon because startups are hard). If you’re solving any of them, I’d love to chat. 1. Data synthesis: AI has become really good both at generating and annotating data. The challenge now is to make sure…

I often hear that evals are the most confusing part of creating LLM AI products. It's a shame b/c IMO, domain-specific evals are the most important part of an AI product! I've written a detailed blog post with real examples on how to do this (1/3) hamel.dev/blog/posts/eva…

Last week, I described four design patterns for AI agentic workflows that I believe will drive significant progress this year: Reflection, Tool use, Planning and Multi-agent collaboration. Instead of having an LLM generate its final output directly, an agentic workflow prompts…

This is the most comprehensive tour of TigerBeetle yet. We put everything in here: - LSM optimizations like moving from RUM to RUMBA, and even - consensus tricks like how a cluster with fsync can rival a single node without fsync (!). Hope you enjoy. infoq.com/presentations/…

TigerBeetle @TigerBeetleDB

a month ago

This is the most comprehensive tour of TigerBeetle yet. We put everything in here: - LSM optimizations like moving from RUM to RUMBA, and even - consensus tricks like how a cluster with fsync can rival a single node without fsync (!). Hope you enjoy. infoq.com/presentations/…

1 7 46 26K 19

I think AI agentic workflows will drive massive AI progress this year — perhaps even more than the next generation of foundation models. This is an important trend, and I urge everyone who works in AI to pay attention to it. Today, we mostly use LLMs in zero-shot mode, prompting…

Excited to share something that we've needed since the early open RLHF days: RewardBench, the first benchmark for reward models. 1. We evaluated 30+ of the currently available RMs (w/ DPO too). 2. We created new datasets covering chat, safety, code, math, etc. We learned a lot.…

keyboard_arrow_left Previous keyboard_arrow_right Next

# automating software engineering In my mind, automating software engineering will look similar to automating driving. E.g. in self-driving the progression of increasing autonomy and higher abstraction looks something like: 1. first the human performs all driving actions…

Cognition @cognition_labs

2 months ago

# automating software engineering In my mind, automating software engineering will look similar to automating driving. E.g. in self-driving the progression of increasing autonomy and higher abstraction looks something like: 1. first the human performs all driving actions…

4K 11K 46K 30.4M 28K

Download Video