frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

RTK reports token savings, but our cost benchmarks disagree

https://quesma.com/blog/does-rtk-make-ai-coding-cheaper/
31•michalwarda•1h ago

Comments

vrighter•1h ago
well yeah.... now you're giving it output it wasn't trained on.
nextaccountic•22m ago
What if the next-gen models are trained on RTK output as well? Then you will actually have less tokens in the context window, and the model won't become confused (which would require more turns, wasting tokens)
aeneas_ory•1h ago
All of these "hacks" are snakeoil and I think deep down we all know. Whether it's caveman, RTK, or whatever other vibe-coded productivity/token cost saving hacks/skills/claude.md.

What I had success with (although benchmarks are older) is to index the codebase with a dedicated local code embedding model. It's a bit expensive on the CPU side but in my benchmarks it reduced token use and wall clock time significantly. Of course, it's always dependent on statistical noise + host system load, and running sufficiently large benchmarks is simply too expensive, so take em with a grain of salt.

Why does it work you may ask? Well, LLMs basically brute force words/phrases and pipe that into find/grep/pgrep/whatever (or as recently discussed here write a python script for it - https://news.ycombinator.com/item?id=49654229). Semantic search looks for similarities so you have to do less brute forcing. Comes of course at the cost of indexing everything first.

You can find the project here: https://github.com/ory/lumen

sreekanth850•1h ago
i don't know if such hacks works, but in C# if you use roslyn mcp, you save a lot.
VulgarExigency•31m ago
I don't think they're comparable. RTK just modifies the output of CLI tools to reduce the number of tokens, a Roslyn MCP gives the agent a fundamentally superior way of interacting with a C# codebase.
sreekanth850•30m ago
Yes. and i find model makes less errors and reasoning the codebase well, especially when you do a large refactor.
antupis•22m ago
My main issue with rtk is that rtk randomly messes modification and agent start polling same tool continuously.
fwlr•54m ago
This makes sense. “Don’t try to penny-pinch your employees” is a lesson most managers learn eventually, and I guess agent-orchestrators will have to learn it too.
gillesjacobs•39m ago
Main takeaway:

``` Average cost per attempt, without → with RTK:

Claude/Fable: $1.72 → $1.64 (~5% cheaper) DeepSeek: $0.115 → $0.121 (~5% more expensive)

Almost all Claude savings came from a single task. Excluding it, savings were under 1%. ```

It took me a few rereads to parse out the top-line. This article really buries the lede.

fleetfox•23m ago
I don't understand how this or all these magic skill bundles and methodologies get traction and why they are so popular. It's either plain worse or has serious trade offs.
semiquaver•16m ago
Just another instance of the bitter lesson. The model itself knows how to be clever and conserve tokens in command output by using shell primitives and as the models get smarter they get better at anticipating large output and defensively adapting the input commands.
hokkos•12m ago
If you are using maven you should tell your agent to use its quiet mode or rtk, because mvn love to write a lot of useless output.

The Waymo effect: how AI is quietly making research less collaborative

https://www.researchagenda.news/articles/the-waymo-effect.html
87•JohnHammersley•1h ago•54 comments

RTK reports token savings, but our cost benchmarks disagree

https://quesma.com/blog/does-rtk-make-ai-coding-cheaper/
34•michalwarda•1h ago•12 comments

Cherenkov Radiation - traveling faster than light

http://www.iaea.org/newscenter/news/what-is-cherenkov-radiation
113•andsoitis•3h ago•63 comments

So you want to use OpenRouter?

https://mmoustafa.com/blog/so-you-want-to-use-openrouter/
233•player85•2d ago•40 comments

Shopify is moving from React Native back to Swift and Kotlin

https://shopify.engineering/back-to-native
1113•fnthawar2•22h ago•813 comments

Claude is no longer available for minors

https://support.claude.com/en/articles/15171100-age-assurance-on-claude
126•Muhammad523•1h ago•167 comments

Instagram's head says engagement falls by half without the algorithm

https://thenextweb.com/news/mosseri-instagram-algorithm-opt-out-engagement-australia
37•hsuduebc2•1h ago•33 comments

Don't let anyone take away your big box of cables

https://blog.jim-nielsen.com/2026/hands-off-my-cables/
599•Brajeshwar•21h ago•372 comments

iPod Classic 6G in QEMU

https://www.reddit.com/r/emulation/s/VL4Au2HGxq
41•dmonterocrespo•2d ago•5 comments

Working with Git Worktrees in Magit

https://emacsredux.com/blog/2026/09/02/working-with-git-worktrees-in-magit/
70•srijan4•3d ago•25 comments

OpenAI Agents API

https://developers.openai.com/api/docs/guides/agents-api/overview
280•aquir•16h ago•157 comments

CSS Curiosities of the Past

https://vale.rocks/posts/css-relics
25•robin_reala•4h ago•12 comments

RISC-V Emulator and Linux System from Scratch

https://github.com/WerWolv/riscv-emulator
8•Bluestein•3d ago•2 comments

Mexican student creates an acoustic fire extinguisher to put out fire in seconds

https://www.upsocl.com/en/16-year-old-mexican-student-creates-an-acoustic-fire-extinguisher-that-...
263•rguiscard•11h ago•85 comments

The Deathray: A simple way for an untrusted site to freeze a Mac

https://auberon.xyz/blog/posts/deathray/
224•auberonedu•16h ago•147 comments

Technique for Manipulating Satellite Photos Now Reveals Ancient Images (2025)

https://spinoff.nasa.gov/Manipulating_Satellite_Photos_Now_Reveals_Ancient_Images
348•gumby•21h ago•56 comments

GPT‑Live‑1 in the API

https://openai.com/index/introducing-gpt-live-1-in-the-api/
46•arittr•6h ago•39 comments

Nine coding harnesses vs. your laptop

https://nasutton.notion.site/Nine-coding-harnesses-vs-your-laptop-3d139990182b80d59fa3cf500f0450b...
124•nasutton12•13h ago•39 comments

Neijuan

https://en.wikipedia.org/wiki/Neijuan
65•vermilingua•4h ago•19 comments

An interactive tour of the spanning tree protocol

https://vincent.bernat.ch/en/blog/2026-spanning-tree
26•zdw•2d ago•2 comments

More questions about whether researchers can trust OpenAI with unpublished math

https://mathstodon.xyz/@andreasthom/117240535270608201
827•pred_•1d ago•762 comments

Music Theory for the 21st-Century Classroom

https://musictheory.pugetsound.edu/mt21c/MusicTheory.html
247•aanet•19h ago•107 comments

Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-Astra

https://cognition.com/blog/swe-2
426•seelos•21h ago•176 comments

Forgejo <=16.0.3 Critical RCE

https://codeberg.org/forgejo/forgejo/src/branch/forgejo/release-notes-published/16.0.4.md
195•weierstass•20h ago•75 comments

Neki – Sharded Postgres

https://planetscale.com/blog/introducing-neki
252•simon_weber•20h ago•133 comments

Proof of Capture: Apple Reference Image, but open source and using steganography

https://merybenavente.me/blog/proof-of-capture
118•merybenavente•16h ago•67 comments

NTSB issues investigative update on B-767 runway excursion accident in Miami

https://www.ntsb.gov:443/news/press-releases/Pages/NR20260909.aspx
117•mckn1ght•15h ago•206 comments

Rust is tier-1 language at Microsoft

https://rustfoundation.org/media/guest-post-rust-is-tier-1-language-at-microsoft/
695•mmastrac•22h ago•450 comments

Douglas Hofstadter: Analogy as the Core of Cognition [video]

https://www.youtube.com/watch?v=n8m7lFQ3njk
192•tosh•4d ago•86 comments

Recursion into madness

https://blog.coredump.cx/p/recursion-into-madness
84•surprisetalk•3d ago•20 comments