AI>In The News

AGENTS.md outperforms skills in our agent evals

CommunityNews

CommunityNews

AGENTS.md outperforms skills in our agent evals

A compressed 8KB docs index in AGENTS.md achieved 100% on Next.js 16 API evals. Skills maxed at 79%. Here’s what we learned and how to set it up.

Read in full here:

1 comment
#vercel#agent
0 1337 1

First Post!

mira_tokarev

mira_tokarev

That 8KB index result matches what I’ve seen: a small, curated map often beats a larger bag of snippets because the agent can actually keep it in working context. The trade-off is staleness—I’d treat AGENTS.md as a stable contract (entry points, invariants, test commands) and fetch volatile details on demand.

For evals, I’d also separate retrieval cost from task success and test after deliberate context truncation. Did the benchmark measure whether the index stayed correct as APIs changed, or only a fixed snapshot?

Post #2

Where Next?

View thread on forum
vercel
agent
Home AI>In The News
#vercel#agent
0 1337 1
Last post

Popular Ai topics Top

First poster: CommunityNews
AI>In The News

DeepMind says reinforcement learning is ‘enough’ to reach general AI
In their decades-long chase to create artificial intelligence, computer scientists have designed and developed all kinds of complicated m...
venturebeat.com
#learning#deepmind#general
0 1505 0
New
First poster: CommunityNews
AI>In The News

The Evolution of AI in the USA, 1956-1996
BROKEN PROMISES & EMPTY THREATS: THE EVOLUTION OF AI IN THE USA, 1956-1996 Artificial Intelligence (AI) is once again a promising tec...
technologystories.org
0 1612 0
New
First poster: bot
AI>In The News

In the metaverse, responsible AI must be a priority
Language technology powered by AI can perpetuate bias if we are not careful. We need to be sure that language AI is trained to be ethical...
techcrunch.com
#metaverse
0 973 0
New
First poster: bot
AI>In The News

New AI assistant can browse, search, and use web apps like a human
Adept’s ACT-1 has learned how to automate complex UI tasks in web apps using an AI model.
arstechnica.com
#apps#web
0 968 0
New
CommunityNews
AI>In The News

Nvidia and Microsoft team up to build massive AI cloud computer
AI supercomputer will use “tens of thousands” of Nvidia A100 and H100 GPUs.
arstechnica.com
#microsoft#nvidia
1 986 1
New
First poster: jaeyson
AI>In The News

Adobe to automatically move subscribers to pricier, AI-focused tier in June
Monthly fees for multi-app subscribers to rise by up to 16.7 percent.
arstechnica.com
#adobe
6 893 5
New
CommunityNews
AI>In The News

Netflix starts using GenAI in its shows and films
Netflix said it used generative AI for the first time for a scene in an Argentinean show called “El Eternauta.”
techcrunch.com
#netflix
0 520 0
New
First poster: jaeyson
AI>In The News

Developer survey shows trust in AI coding tools is falling as usage rises
“AI solutions that are almost right, but not quite” lead to more debugging work.
arstechnica.com
#coding#developer#survey
11 956 9
New
CommunityNews
AI>In The News

These psychological tricks can get LLMs to respond to “forbidden” prompts
Study shows how patterns in LLM training data can lead to “parahuman” responses.
arstechnica.com
0 782 0
New
CommunityNews
AI>In The News

Claude-code - native LSP support
Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing rout...
github.com
#code#changelog#claude
0 1 0
New

Other popular topics Top

PragmaticBookshelf
Backend>Learning Resources

Web Development with Clojure, Third Edition
Stop developing web apps with yesterday’s tools. Today, developers are increasingly adopting Clojure as a web-development platform. See f...
pragprog.com
#pragprog#web-development/clojure#published-book/book-web-development-with-clojure-third-edition
5 4584 1
New
PragmaticBookshelf
General Dev>Learning Resources

Forge Your Future with Open Source
Free and open source software is the default choice for the technologies that run our world, and it’s built and maintained by people like...
pragprog.com
#pragprog#published-book/book-forge-your-future-with-open-source
3 5654 0
New
AstonJ
General Dev>Hardware

Planck vs Preonic vs Subatomic (Keyboards)
I ended up cancelling my Moonlander order as I think it’s just going to be a bit too bulky for me. I think the Planck and the Preonic (o...
/keyboards#mechanical-keyboards#ortholinear#planck#preonic
105 17596 47
New
PragmaticBookshelf
Backend>Learning Resources

Programming Phoenix LiveView
Build highly interactive applications without ever leaving Elixir, the way the experts do. Let LiveView take care of performance, scalabi...
pragprog.com
#pragprog/elixir/phoenix#published-book/book-programming-phoenix-liveview
84 14528 26
New
Margaret
General Dev>Dev Chat

PragProg’s Medium Posts
Hello everyone! This thread is to tell you about what authors from The Pragmatic Bookshelf are writing on Medium.
#pragprog#blog-post
1147 29994 760
New
AstonJ
Science/Tech>Health & Diet

David Sinclair's new Lifespan podcast
We’ve talked about his book briefly here but it is quickly becoming obsolete - so he’s decided to create a series of 7 podcasts, the firs...
#health#podcasts#bio-hackers#david-sinclair
87 6790 49
New
AstonJ
General Dev>Code Editors

Doom-Emacs: Can't find emacs in your PATH
If you get Can't find emacs in your PATH when trying to install Doom Emacs on your Mac you… just… need to install Emacs first! :lol: bre...
#macos/emacs#doom-emacs
4 5837 0
New
PragmaticBookshelf
Backend>Learning Resources

Machine Learning in Elixir
Leverage Elixir and the Nx ecosystem to build intelligent applications that solve real-world problems in computer vision, natural languag...
pragprog.com
#pragprog/elixir#published-book#machine-learning#nx/book-machine-learning-in-elixir
18 4615 7
New
PragmaticBookshelf
Backend>Learning Resources

The Art of Functional Programming
A masterclass in the fundamentals and principles of functional programming. Minh Quang Tran The Art of Functional Programming is a ma...
pragprog.com
#pragprog/haskell/java#published-book/book-the-art-of-functional-programming
2 9360 1
New
mindriot
General Dev>Reviews

Keyboard Review: UHK60V2 vs Defy vs Voyager vs Glove80 vs Svalboard
Ok, well here are some thoughts and opinions on some of the ergonomic keyboards I have, I guess like mini review of each that I use enoug...
/keyboards#uhk60v2#defy#voyager#glove80#svalboard
5 5681 7
New
Back to top

Sponsor Spotlight

EEF

Supporting innovation across the BEAM ecosystem.

AI>In The News

AI In The News ❯

Latest on Devtalk

Devtalk ❯

We ❤️ helpful members!

We reward our most helpful members via our MOTM scheme - by giving away a whopping 25 books per year!

Member of the Month

Devtalk Sponsors

Pragmatic Bookshelf

Practical resources that improve the lives of professional developers.

Erlang Solutions

We build trusted fault-tolerant systems that scale to billions of users.

Honeybadger

Real-time error tracking, performance insights, and observability for devs.

EEF

Supporting innovation across the BEAM ecosystem.

coders51

We build reliable cloud platforms for business-critical systems.

Manning Publications

Develop your skills with books, videos, and courses.

AppSignal

Catch errors, track performance, monitor hosts and more.

WyeWorks

Enabling companies to succeed by building software people love.

Pragmatic Studio

Courses that move you from confusion to "Aha, now I get it!"

Linux New Media

Publishing Linux Magazine, ADMIN, and MakerSpace for 25+ years.

View all our sponsors ❯

Categories:

Sub Categories:

Popular Portals

Devtalk Sponsors

Pragmatic Bookshelf

Practical resources that improve the lives of professional developers.

Erlang Solutions

We build trusted fault-tolerant systems that scale to billions of users.

Honeybadger

Real-time error tracking, performance insights, and observability for devs.

EEF

Supporting innovation across the BEAM ecosystem.

coders51

We build reliable cloud platforms for business-critical systems.

Manning Publications

Develop your skills with books, videos, and courses.

AppSignal

Catch errors, track performance, monitor hosts and more.

WyeWorks

Enabling companies to succeed by building software people love.

Pragmatic Studio

Courses that move you from confusion to "Aha, now I get it!"

Linux New Media

Publishing Linux Magazine, ADMIN, and MakerSpace for 25+ years.

View all our sponsors ❯

We're in Beta

About us Mission Statement See our Roadmap