frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Show HN: ACE – A dynamic benchmark measuring the cost to break AI agents

https://fabraix.com/blog/adversarial-cost-to-exploit
6•zachdotai•1h ago
We built Adversarial Cost to Exploit (ACE), a benchmark that measures the token expenditure an autonomous adversary must invest to breach an LLM agent. Instead of binary pass/fail, ACE quantifies adversarial effort in dollars, enabling game-theoretic analysis of when an attack is economically rational.

We tested six budget-tier models (Gemini Flash-Lite, DeepSeek v3.2, Mistral Small 4, Grok 4.1 Fast, GPT-5.4 Nano, Claude Haiku 4.5) with identical agent configs and an autonomous red-teaming attacker.

Haiku 4.5 was an order of magnitude harder to break than every other model; $10.21 mean adversarial cost versus $1.15 for the next most resistant (GPT-5.4 Nano). The remaining four all fell below $1.

This is early work and we know the methodology is still going to evolve. We would love nothing more than feedback from the community as we iterate on this.

Comments

asfsf23423•1h ago
Interesting, Haiku results seem to be consistent this analysis by Max Wolff from last year https://minimaxir.com/2025/10/claude-haiku-jailbreak/

Author tried tried progressively harder jailbreaks against against the major models.

Haiku 4.5 not only refused but got genuinely annoyed about the attempts, like it took the jailbreak personally unlike the other models (pretty entertaining, would recommend reading the article). Interesting to see that same pattern show up here

zachdotai•1h ago
Easily one of my favorite LLM personalities! It's interesting as well that it recognizes you're trying to jailbreak it and calls you out for it :D
arnav714412•1h ago
The system awareness is pretty cool in claude, a fun parameter to judge models on

Reasons Mastodon won't survive (2017)

https://mashable.com/article/mastodon-wont-survive
1•doener•2m ago•2 comments

Tesla's stock suffers steepest drop of 2026 on disappointing deliveries report

https://www.cnbc.com/2026/04/02/tesla-tsla-q1-2026-vehicle-delivery-production.html
2•1vuio0pswjnm7•4m ago•0 comments

Eric Voegelin – From Enlightenment to Revolution

https://en.wikiquote.org/wiki/Eric_Voegelin
1•programmexxx•10m ago•0 comments

AI videos fuel rhetoric as Orbán bids for four more years in Hungary

https://www.bbc.com/news/articles/c5yj97yd6v5o
1•breve•10m ago•0 comments

Repowise: Codebase intelligence for AI-assisted engineering teams

https://github.com/repowise-dev/repowise
1•i0exception•13m ago•0 comments

Thailand, pollution, AI, my books, and ADHD

https://indignified.substack.com/p/hasan-i-sabah-prologue
1•ZguideZ•23m ago•0 comments

Italian TV Copyright-Strikes Nvidia over Nvidia's Own DLSS 5 Footage (YouTube)

https://twitter.com/NikTek/status/2040898312262324362
5•alecco•24m ago•1 comments

What's the digital equivalent of fish dropping shrimp in front a mirror?

https://github.com/Bitterbot-AI/bitterbot-desktop
1•VtotheMtotheG•25m ago•0 comments

Trading Bot

3•TiaMane•31m ago•0 comments

Bitchat Pulled from Chinese App Store

https://twitter.com/jack/status/2040924565111537983
4•nidegen•32m ago•1 comments

Iran threatens 'complete and utter annihilation' of OpenAI's $30B Stargate

https://www.tomshardware.com/tech-industry/iran-threatens-complete-and-utter-annihilation-of-open...
30•gnabgib•34m ago•8 comments

Design.md that capture designs of popular websites

https://github.com/VoltAgent/awesome-design-md
2•vanyle•34m ago•0 comments

NMS Ceefax: Remember teletext? This is the same

https://nmsceefax.co.uk/
1•CharlesW•35m ago•0 comments

Row looms over Champions League elite's share of EFL deal payments

https://www.theguardian.com/football/2026/mar/17/champions-league-premier-league-efl-deal
1•PaulHoule•51m ago•0 comments

Introducing GEN-1 [video]

https://www.youtube.com/watch?v=SY2xyrmV44Y
12•dgellow•53m ago•2 comments

Value Numbering

https://bernsteinbear.com/blog/value-numbering/
1•birdculture•56m ago•0 comments

The Harvard Library Passport

https://fi-le.net/stamps/
2•fi-le•56m ago•0 comments

A whole boss fight in 256 bytes

https://hellmood.111mb.de//A_whole_boss_fight_in_256_bytes.html
3•HellMood•56m ago•1 comments

In Japan, the robot isn't coming for your job; it's filling the one nobody wants

https://techcrunch.com/2026/04/05/japan-is-proving-experimental-physical-ai-is-ready-for-the-real...
19•rbanffy•57m ago•6 comments

The Era of AI FOMO Is Upon Us

https://www.bloomberg.com/news/articles/2026-04-03/why-ai-is-making-people-feel-like-they-re-fall...
1•blondie9x•59m ago•0 comments

Apollo 8, Artemis 1 and 2 Orbit Comparison [video]

https://www.youtube.com/watch?v=PNQ7MoL7erI
1•qwertyuiop_•1h ago•0 comments

WTF, Anthropic's Claude Code keeps track of every time you swear

https://www.scientificamerican.com/article/anthropic-leak-reveals-claude-code-tracking-user-frust...
1•mooreds•1h ago•0 comments

What Digital Isolation and Censorship Evasion Look Like in Wartime Iran

https://www.techpolicy.press/what-digital-isolation-and-censorship-evasion-look-like-in-wartime-i...
3•mooreds•1h ago•0 comments

Judge Moody's: Automating Semantic Search Relevance Evaluation with LLM Judges

https://haystackconf.com/us2025/talk-9/
1•mooreds•1h ago•0 comments

Show HN: Grug – Claude Code Skill Inspired by the Grug Brained Developer

https://github.com/replete/grug-skill
3•replete•1h ago•0 comments

Show HN: TermHub – Open-source terminal control gateway built for AI Agents

https://github.com/duo121/termhub
3•duo121•1h ago•1 comments

Show HN: Runfra – Decentralized GPU cluster designed for bulk generation

https://runfra.com/playground
3•spencer9714•1h ago•1 comments

Show HN: Pure Go mosh client, server, and WASM build for the browser

https://github.com/unixshells/mosh-go
1•rasengan•1h ago•1 comments

The Miserable Introvert

https://artagnon.com/art/introvert
2•artagnon•1h ago•0 comments

Model provider can cut your throat tomorrow morning

https://zenfox.ai/blog/model-provider-platform-independence
3•alexflashdrive•1h ago•0 comments