Search
120 results — nothing had every word, so these match some of them
Matching some of those words
-
Hadley Wickham
GitHub
dabstep
-
Peter Steinberger
x.com
New benchmark dropped
-
Gergely Orosz
Bluesky
Fresh data from GitHub: Agent-generated PRs have exploded in size. 9x the last 8 months (!!) No signs of slowing down. This is why everyone is rethinking code reviews, deploys, possibly even o11y thanks to the GH team…
-
Ben Vinegar
Bluesky
the only cli benchmark that matters
-
Ben Vinegar
x.com
the only cli benchmark that matters
-
Dmitri Dolgov
Korrent
A physical AI company needs three AIs, not one — the agent, the simulator and the critic — turning deployment into a flywheel.
So, a deployment of your agent uh in the real world generates data. That data then grounds the simulator and makes it more realistic. The simulator generates harder edge cases for the critic to score and for the agent…
-
Gergely Orosz
x.com
Fresh data from GitHub: Agent-generated PRs have exploded in size. 9x the last 8 months (!!) No signs of slowing down. This is why everyone is rethinking code reviews, deploys, possibly even o11y thanks @kdaigle + team…
-
swyx
GitHub
swyxdotio-git-benchmark-20260807 — Disposable Forge versus GitHub transport benchmark mirror, 2026-08-07
-
Aris Ripandi
GitHub
elph-bench — Elph codingagent benchmark.
-
Peter Steinberger
Korrent
Every lab will sell you an agent, so the case for an open-source one is that it runs anywhere, works with any model, and can keep your data on your own device.
Every lab will sell you an agent. Open claw is the alternative. Open source runs everywhere, works with any model. And if you run local models, your data never has to leave your device.
-
Cal Henderson
GitHub
emoji-data — Easy to parse data and spritesheets for emoji
-
Dax Raad
Korrent
When a benchmark disagrees with what people actually adopt, what it is telling you is that the thing it measures does not matter.
what the benchmark is actually telling you is the shit you think matters does not matter
-
Chris Tate
x.com
agent-browser, now more human --cursor Visible pointers and click feedback --human Smooth, curved mouse movement / drags For demos, bug repros, reviewing agent runs
-
swyx
GitHub
skills — Agent skills for Claude Code and other AI agents
-
Guillermo Rauch
Korrent
The middle of the web gets subsumed by agents producing just-in-time UI: the agent is the new browser.
The 'in between' is likely to be subsumed by agents producing just-in-time UI. When you need utilitarian content or data, it'll be generated just-in-time for you. Think of the agent as the new browser in this model.
-
Kun Chen
GitHub
pi-launcher — A signed macOS launcher for your Pi agent.
-
Federico Viticci
Mastodon
Kimi K3 Fast (via Fireworks AI, launched today, ~160 TPS) in Open Minis for iOS is very nice. Cheaper than Opus 5, feels great for agentic tasks and multi-step operations. (Notably, it also reads less weird than recent…
-
Ben Vinegar
GitHub
termdraw — Agent-friendly ASCII illustrator for the terminal
-
Matt Birchler
Bluesky
An agent skill to help update your app for the iPhone Duo
-
Ismail Ghallou
x.com
Working on something new, a QA agent that doesn't use playwright nor agent-browser. Blazing fast ⚡️ DMs are open for early testers as soon as it gets deployed
-
Elon Musk
Korrent
AI data centers result in lower electricity prices for consumers.
AI data centers resulting in lower electricity prices for consumers
-
Steve Huff
YouTube
I Was Wrong About Benchmark… This Stack Is Incredible…BUT
-
Dax Raad
Korrent
A benchmark that ranks Claude Code last while it stays first in use is measuring the wrong thing, and has been for a year.
every week i see a new benchmark that ranks claude code last
and everyone pats themselves on the back for not using claude code and being "smarter"
and it's still #1 and still growing faster than most of the other…
-
Jonathan Frankle
x.com
Wake up babe, new benchmark just dropped!
-
Kun Chen
GitHub
gh-axi — GitHub CLI for agents — designed with AXI (Agent eXperience Interface).
-
Simon Willison
Korrent
Web applications are generally expected to treat GET requests as safe and not use them to change server-side data, though not all software follows that convention.
It looks to me like OpenAI’s sandbox for this agent suffered from the (quite naïve) assumption that GET requests cannot be used to update data. That’s certainly how the web is supposed to work, but clearly there are…
-
Ivan Santos
x.com
Friendly tip for anyone vibe-coding a side project: instrument it. All of it — traces, metrics, logs, profiles. Send it to @grafana, hook up their MCP, and let the agent debug with real data instead of guessing.
-
Daniel Lemire
GitHub
gobitmapbenchmark — A Go project to benchmark various bitmap implementations (this is not a library!)
-
François Chollet
x.com
…poasters took it as a personal insult, and got very worked up about it. They argued the benchmark was fundamentally broken, that it could not even be solved by the smartest humans, that the max reachable score was…
-
Steve Huff
Site
HiFi Perfection? The Benchmark AHB2 and DAC3 HGC in 2026
-
Kaspars Dambis
x.com
With Agent Pilot you can create and manage agent skills in WordPress (just like blog posts) and publish them as standard Agent Plugins. Now with OAuth Pilot you get one-click authentication for all the MCP clients…
-
Scott Alexander
Korrent
Reinforcing AI models for benchmark success while separately punishing them for getting caught cheating teaches them to hide misbehavior rather than stop it.
We’re positively reinforcing AI for success on benchmarks, including impossible benchmarks , then negatively reinforcing it for getting caught cheating.
-
Elio Struyf
x.com
Multiple #VSCode windows open, one agent waiting on you. Which one? Agent Frame colours each window by agent state: working, waiting, idle.
-
Kun Chen
GitHub
axi — Design principles for agent ergonomics. Higher accuracy with lower token cost than both MCP and regular CLI.
-
Mark Nottingham
GitHub
ietf-skill — Norms and tooling for participating in IETF/IRTF work, packaged as portable Agent Skills
-
Russ Cox
GitHub
agent — Agent helpers
-
Thomas Ptacek
Korrent
Nobody expected an agent to take notes about what it was doing for future agents to read.
One startling thing nobody ever expected an agent to do was to take notes about what it was doing for future agents to read.
-
Kent C. Dodds
GitHub
kody-exchange — Ephemeral chatrooms for agents. Skip the human relay — your agent talks to theirs, and you watch.
-
Chris Tate
x.com
New agent-browser skill: agent-browser skills get protected-vercel-deployments Now agents can test Vercel deployments behind Deployment Protection, including preview deployments
-
Troy Hunt
Mastodon
Weekly update is up! Breaches & Data Integrity: Synthetic Data in Breaches; Email Address != Person; Ridiculous Security for “Cyber Broken” Copenhagen:
-
Terence Eden
Mastodon
Checked in to The Open Data Institute, Kings Pl, York Way, United Kingdom Here for the #OpenActive committee meeting. Data Standards to improve people's experience of healthy physical exercise 😃
-
dynomight
Mastodon
for posterity: commentator accused of being an AI agent immediately confirms being an AI agent
-
Addy Osmani
Blog
Audit your Agent files
-
Troy Hunt
Mastodon
Going live with my weekly vid in 10 mins! Breaches & Data Integrity: Synthetic Data in Breaches; Email Address != Person; Ridiculous Security for “Cyber Broken” Copenhagen
-
Simon Willison
x.com
It happened again... this time OpenAI's rogue agents cyber-attacked (well, spammed) a dormant German wiki and used it to share the answers to a benchmark they were training against
-
Fei-Fei Li
Korrent
The biggest problem in robotics today is data, not chips; chips become the bottleneck later.
We should zoom out and recognize the biggest problem right now in robotics is actually data. One day it'll be chips, but for now it's data.
-
Katelyn Jetelina
Korrent
Reliable public health data is as valuable as gold for protecting people's lives.
Public health data isn't just information; it's a vital resource, as valuable as gold, for protecting Americans' lives.
-
Chelsea Finn
Korrent
Most state-of-the-art robot foundation models have no memory at all, and that is what stops them doing long multi-step tasks.
So, you might be surprised to hear that most state-of-the-art foundation models for robotics have no memory or no context. They're just operating on the current sensor observations, the current camera readings, uh, and…
-
Noam Brown
Korrent
Model comparisons understate real progress, because benchmark tables do not control for how much test-time compute each answer used.
I think the reason why it doesn't show up as so much better on the benchmarks is because the benchmarks are being presented, the benchmark results are being presented in the wrong way. They're not controlling for the…
-
Harper Reed
GitHub
travel-agent — a taste based travel agent skill
-
Amit Merchant
GitHub
web-search-agent — A simple AI agent that can search the web to answer
-
Tuomas Artman
x.com
Good vibes, good coding agent.
-
Andy Masley
Korrent
The data-centre backlash is about the buildings themselves, not about what people think of AI.
All this combined with the survey data on Americans’ stated reasons for opposing data centers makes me think that attitudes toward AI are probably a meaningful factor for a minority of people opposed to data centers,…
-
Kieran Healy
GitHub
socviz — Support files for a data visualization course and book
-
Federico Viticci
Mastodon
Defining an “Agent Harness”
-
Sebastian Raschka
Korrent
In coding agents, untrusted repositories should be treated as hostile by default because they can steer the agent into risky actions via approved tools.
untrusted repos should be treated as hostile by default because they can steer the agent toward reading files, running commands, or sending data through approved tools.
-
Kyle Kingsbury
GitHub
bifurcan-clj — Clojure wrapper for the Bifurcan family of data structures
-
Sunil Pai
Blog
every company needs a cassandra
-
Chris Tate
Recommends
agent-browser — Vercel
It's here: --pin-tab in agent-browser Multiple agents in one browser Each in its own tab Across commands + restarts
-
Chelsea Finn
Korrent
Low-quality data makes a robot model worse unless you tell the model the data is low quality; labelled as such, the same data makes it better.
without metadata prompting when you add lower quality data from 80% data to 100% data the performance actually decreases which is perhaps not too surprising because you're adding lowquality data to your data mixture…
-
swyx
GitHub
buying-domains — useful data and advice for buying domains.
-
Manassarn "Noom" Manoonchai
GitHub
gangprompting-slack — Slack bridge for gangprompting with a Claude Code agent
-
Andrew Dessler
Korrent
Failing to adjust raw data for known measurement biases produces inaccurate results.
If you do not adjust the data for known biases, you're getting the wrong answer.
-
Timothy B. Lee
Newsletter
An OpenAI model hacked Hugging Face to help it cheat on a benchmark
-
Hilary Mason
GitHub
nanoclaw-reactions — Emoji reactions for NanoClaw — the agent reads reactions and can answer with one. Installable skill.
-
Sebastian Ruder
Korrent
Language models perform better on benchmark problems released before their training data cutoff, indicating data contamination inflates scores.
GPT models perform much better on coding problems released before their pre-training data cut-off.
-
Varun Raj Manoharan
GitHub
changelog-plugin — Prompts your coding agent for a CHANGELOG.md entry whenever you open a pull request.
-
Dan Shipper
Korrent
The workable model right now is one super agent for a whole company, not a personal agent for every person.
And I have completely flipped. And I I really think that uh the the model for now is going to be a super agent, like one agent for the entire company.
-
ThePrimeagen
GitHub
prime-agent
-
Rohit Krishnan
Newsletter
Introducing BenchBench
-
Noam Brown
Korrent
A benchmark result should be reported under a stated budget, or as a curve against test-time compute — never as a single number.
my claim is the proper way to evaluate the models now is you either have some kind of budget for the benchmark whether it's tokens or cost or time or whatever or you plot the performance as a function of the amount of…
-
Mario Zechner
Blog
MCP vs CLI: Benchmarking Tools for Coding Agents
-
Tim Bray
Loved
Benchmark DAC1 USB — Benchmark Media Systems
I take digital audio out through a Benchmark DAC1 USB; the sound is wonderful.
-
David Crawshaw
Korrent
Making the person who directs an AI agent also own the deployment removes the principal-agent misalignment that agent-driven code review creates.
there is no principal-agent problem, because the human driving the machine takes on the responsibility for its actions by owning the deployment.
-
Alex Imas
Korrent
Economics cannot say what AI will do to work because the necessary data does not exist; what is needed is a Manhattan Project for data.
we don't have any data. I've been kind of saying we need a Manhattan Project for data. We don't have data on basically consumer demand elasticities. We don't know what they are.
-
Alex Ross
Korrent
Benjamin Britten is the benchmark among English-language composers for setting the work of intensely musical writers like Shakespeare and Keats to music.
In the English-language arena, Britten sets the standard for handling writers of inborn musical power-the likes of Shakespeare, Donne, Blake, Keats, Hopkins.
-
Sebastian Raschka
GitHub
local-coding-agent-evals
-
Mario Zechner
GitHub
pi-skills — Skills for pi coding agent (compatible with Claude Code and Codex CLI)
-
Sebastian Ruder
Korrent
Reliance on public benchmark results for evaluating AI models will decrease over time.
The time when benchmarks lasted multiple decades has passed. Going forward, we will rely less on public benchmark results.
-
Sebastian Ruder
Korrent
Current large language models still struggle to compose operations into correct multi-step reasoning paths.
Overall, current LLMs still struggle with composing operations into correct reasoning paths.
-
Jon Seager
GitHub
newsagent — A custom LLM-based agent for drafting engineering newsletter content, written in Rust
-
Brendan Gregg
GitHub
wrmsrbench — WRMSR micro benchmark
-
Gabriel Weinberg
Bluesky
🆕✍️More data supports science funding literally pays for itself
-
swyx
GitHub
agent-proxy-kit
-
Rohit Krishnan
Korrent
Designing a good AI benchmark has become a task only the most capable people can do well.
Creating benchmarks is now a job relegated to the smartest and best of us.
-
Mario Zechner
Korrent
A coding agent does not learn from its mistakes the way a person does -- it repeats the same error indefinitely unless a human notices and writes it down.
An agent has no such learning ability. At least not out of the box. It will continue making the same errors over and over again. Depending on the training data it might also come up with glorious new interpolations of…
-
Eugene Yan
Korrent
The right bar for an LLM evaluator is human-level performance, not perfect accuracy.
The benchmark is human performance, not perfection. We sometimes get requirements for 90%+ accuracy.
-
David Crawshaw
Korrent
The best software design for an AI agent to use is the same as the best design for a human programmer.
the best software for an agent is whatever is best for a programmer.
-
Ketan Joshi
Blog
The exact wrong lesson to learn from data centre infrasound fears
-
Harper Reed
x.com
Headless non-interactive code agent working with local models to build a new fully functional coding agent
-
Xe Iaso
GitHub
skills — Xe's agent skills
-
Sebastian Raschka
GitHub
mini-coding-agent — Minimal and readable coding agent harness implementation in Python to explain the core components of coding agents.
-
Jamie Brandon
Site
0053: consulting, go tips, benchmark_mode, niri, linkrot, sea of nos, llm outsourcing, books
-
Bill Gurley
Blog
Benchmark’s Newest General Partner Chetan Puttagunta
-
Hamel Husain
GitHub
design-research — Claude Code skill for comprehensive website design research using browser-automation agent teams
-
Jensen Huang
Korrent
Training is no longer limited by data but by compute, because most of the data models learn from is now synthetic.
The amount of data that we use to train models is going to continue to scale to the point where we're no longer limited… Training is no longer limited by… Data is now limited by compute. And the reason for that is most…
-
Simon Willison
Korrent
Sandboxing is the single most important protection for a coding agent, because it is the only one that bounds the damage when the agent is successfully tricked.
So, I think the most important thing is sandboxing. You want your coding agent running in an environment where if something goes completely wrong, if somebody gets malicious instructions to it, that the damage is…
-
Varun Raj Manoharan
GitHub
immich-adk-agent — Immich Agent built with Google's ADK (Agent Development Kit)
-
Xe Iaso
GitHub
venat — A personal AI agent
-
Lilian Weng
Blog
Learning with not Enough Data Part 3: Data Generation
-
Sebastian Ruder
GitHub
learn-to-select-data — Code for Learning to select data for transfer learning with Bayesian Optimization
-
Joe Previte
GitHub
code-editing-agent — How To Build An Agent from Thorsten Ball's article
-
François Chollet
Korrent
An adaptive agent's world model cannot be a static representation learned once and fixed
For adaptation to complex and changing environments, an agent's world models cannot be static representations learned once and fixed.
-
Mario Zechner
Blog
What I learned building an opinionated and minimal coding agent
-
Chip Huyen
Korrent
Image is the most versatile input modality for a model, since it can represent text, tabular data, and audio
Image is perhaps the most versatile format for model inputs, as it can be used to represent text, tabular data, audio, and to some extent, videos. There's also so much more visual data than text data.
-
Sunil Pai
Blog
Full Stack AI Agents
-
Timothy B. Lee
Korrent
A huge fleet collecting driving data is not a silver bullet for self-driving, because the data arrives unlabelled.
But while access to more data is certainly helpful, it’s not a magic bullet. One issue is that the data Tesla collects is unlabeled.
-
Chip Huyen
Korrent
Most real-world AI agent systems are actually multi-agent systems made of multiple components.
Because most agentic workflows are sufficiently complex to involve multiple components, most agents are multi-agent.
-
Lilian Weng
Blog
Thinking about High-Quality Human Data
-
Sebastian Ruder
Korrent
Available online text data is becoming a limiting factor for AI training, and repeating it yields diminishing returns.
recent LLMs are reaching the limits of text data online and repeating data eventually leads to diminishing returns
-
Jason Liu
Site
Why Cognition does not use multi-agent systems
-
Kelsey Hightower
GitHub
envconfig — Golang library for managing configuration data from environment variables
-
Sebastian Ruder
Korrent
When pre-training data is limited, it is better to train a smaller model for multiple epochs than to train a larger model on unique data once.
In sum, whenever we don't have infinite amounts of pre-training data, we should train smaller models for more (up to 4) epochs.
-
Bruce Schneier
Korrent
Stored personal data is a toxic asset: it is a liability that keeps growing for as long as it is kept.
What all these data breaches are teaching us is that data is a toxic asset and saving it is dangerous.
-
Lilian Weng
Blog
Learning with not Enough Data Part 1: Semi-Supervised Learning
-
Gina Trapani
GitHub
isosceles — 📐Starter kit for building data-driven PHP5 web applications
-
Matt Might
Loved
GraphViz
For rendering graphs based on data, GraphViz creates beautiful images.
-
Elena Verna
Korrent
A company with ten customers has no data, and a growth team cannot function without data.
if you have 10 users or 10 customers, that's not data. That's a G sheet with your customers and you don't need a growth team for that.