Hacker News Digest
Saturday, May 2, 2026
In This Issue
- Hacker News
- The Gay Jailbreak Technique
- Ti-84 Evo
- City Learns Flock Accessed Cameras in Children's Gymnastics Room as a Sales Demo
- Uber Torches 2026 AI Budget on Claude Code in Four Months
- AI Uses Less Water Than the Public Thinks
- New research suggests people can communicate and practice skills while dreaming
- Meta abandons open-source Llama for proprietary Muse Spark
- Show HN: Spec27 – Spec-driven validation for AI agents
- Show HN: Omar – A TUI for managing 100 coding agents
Zipper Data Brief
May 02, 2026
Your daily digest of the best from Hacker News
Top 6 Trending
#1
535 points
· bobsmooth
· comments
Discussion Summary
The discussion centers on a jailbreak technique that exploits LLM guardrails by using casual, gay-coded language to bypass safety filters, with commenters debating whether the technique actually works on current models, how guardrails function, and the broader absurdity of prompt injection vulnerabilities in AI systems.
The discussion centers on a jailbreak technique that exploits LLM guardrails by using casual, gay-coded language to bypass safety filters, with commenters debating whether the technique actually works on current models, how guardrails function, and the broader absurdity of prompt injection vulnerabilities in AI systems.
#2
473 points
· thatxliner
· comments
Discussion Summary
The TI-84 Evo is overpriced and relies on educational lock-in rather than genuine innovation, with commenters nostalgic about programming their calculators but skeptical that the device justifies its cost when cheaper alternatives and software solutions exist. Most users found the calculators' real value in learning to program rather than performing calculations, and many view TI's market dominance as exploitative rather than educational.
The TI-84 Evo is overpriced and relies on educational lock-in rather than genuine innovation, with commenters nostalgic about programming their calculators but skeptical that the device justifies its cost when cheaper alternatives and software solutions exist. Most users found the calculators' real value in learning to program rather than performing calculations, and many view TI's market dominance as exploitative rather than educational.
#3
411 points
· joshcsimmons
· comments
Discussion Summary
Flock Safety employees accessed cameras in a children's gymnastics room at a Jewish community center under the guise of a product demo, raising serious questions about surveillance oversight, lack of proper demo environments, notification of affected parties, and evidence suggesting employees may have intentionally browsed sensitive areas rather than conducting legitimate demonstrations.
Flock Safety employees accessed cameras in a children's gymnastics room at a Jewish community center under the guise of a product demo, raising serious questions about surveillance oversight, lack of proper demo environments, notification of affected parties, and evidence suggesting employees may have intentionally browsed sensitive areas rather than conducting legitimate demonstrations.
#4
383 points
· lwhsiao
· comments
Discussion Summary
The discussion overwhelmingly expresses skepticism that Uber's massive AI spending ($6.8M+ in 4 months) demonstrates actual productivity gains, with commenters challenging the equivalence between high token consumption and business value, and noting the lack of concrete outcomes or ROI metrics in the reporting.
The discussion overwhelmingly expresses skepticism that Uber's massive AI spending ($6.8M+ in 4 months) demonstrates actual productivity gains, with commenters challenging the equivalence between high token consumption and business value, and noting the lack of concrete outcomes or ROI metrics in the reporting.
#5
377 points
· hirpslop
· comments
Discussion Summary
The discussion reveals deep disagreement about AI water usage, with commenters questioning the article's methodology (using AI to estimate AI's impact), pointing out regional variations matter more than blanket comparisons, and noting that context about cooling system types and comparison baselines significantly affect whether AI water use is actually concerning.
The discussion reveals deep disagreement about AI water usage, with commenters questioning the article's methodology (using AI to estimate AI's impact), pointing out regional variations matter more than blanket comparisons, and noting that context about cooling system types and comparison baselines significantly affect whether AI water use is actually concerning.
#6
360 points
· XzetaU8
· comments
Discussion Summary
HackerNews users share anecdotal experiences of solving problems and learning skills through dreams, while skeptics point out the research lacks rigorous controls and quantitative evidence, making claims of sleep learning largely unproven despite the intriguing personal accounts.
HackerNews users share anecdotal experiences of solving problems and learning skills through dreams, while skeptics point out the research lacks rigorous controls and quantitative evidence, making claims of sleep learning largely unproven despite the intriguing personal accounts.
AI / Machine Learning
14 points
· CrankyBear
· comments
Discussion Summary
The commenter warns that the article is AI-generated, lacks substantive information, and contains errors such as recommending llama.cpp as an alternative to Llama models themselves.
The commenter warns that the article is AI-generated, lacks substantive information, and contains errors such as recommending llama.cpp as an alternative to Llama models themselves.
13 points
· njyx
· comments
Discussion Summary
Spec27 team introduces their spec-driven validation framework for AI agents, with team members available to discuss the judge system, technical challenges, and safety concerns—while users express interest in reducing hallucinations and improving agent reliability.
Spec27 team introduces their spec-driven validation framework for AI agents, with team members available to discuss the judge system, technical challenges, and safety concerns—while users express interest in reducing hallucinations and improving agent reliability.
Tell HN: Claude Opus 4.7 quota suddenly changed to 0 TPM in Bedrock
12 points
· sarathyweb
· comments
Discussion Summary
Users report that AWS Bedrock suddenly zeroed out Claude Opus 4.7 quota limits, with concerns about unexplained token counting issues and a pattern of unreliable service for production workloads, leading to recommendations to switch providers.
Users report that AWS Bedrock suddenly zeroed out Claude Opus 4.7 quota limits, with concerns about unexplained token counting issues and a pattern of unreliable service for production workloads, leading to recommendations to switch providers.
14 points
· karim7
· comments
Discussion Summary
The creator shared Omar, a TUI tool for managing multiple coding agents, and received congratulatory feedback from the community.
The creator shared Omar, a TUI tool for managing multiple coding agents, and received congratulatory feedback from the community.
41 points
· xodn348
· comments
Discussion Summary
The discussion is mixed, with some commenters finding the fortune-teller skill entertaining while others raise concerns about LLM hallucinations, potential harms from life advice, legal liability, and whether such a feature is a good use of AI resources.
The discussion is mixed, with some commenters finding the fortune-teller skill entertaining while others raise concerns about LLM hallucinations, potential harms from life advice, legal liability, and whether such a feature is a good use of AI resources.
Startups / Business
I'm Peter Roberts, immigration attorney who does work for YC and startups. AMA
169 points
· proberts
· comments
Discussion Summary
A immigration attorney answers detailed questions about U.S. visa and green card processes for tech workers, covering PERM labor certification, H-1B fees, O-1 visas, employment-based green cards, and various other immigration pathways and their practical complications.
A immigration attorney answers detailed questions about U.S. visa and green card processes for tech workers, covering PERM labor certification, H-1B fees, O-1 visas, employment-based green cards, and various other immigration pathways and their practical complications.
12 points
· bassoonspinach
· comments
Discussion Summary
No comments available.
No comments available.
11 points
· patillacode
· comments
Discussion Summary
Users raised concerns about the app's name being incorrect in Spanish, a non-functional demo with invalid credentials, and questioned what other details might have been overlooked during development.
Users raised concerns about the app's name being incorrect in Spanish, a non-functional demo with invalid credentials, and questioned what other details might have been overlooked during development.
6 points
· mooreds
· comments
Discussion Summary
The commenter acknowledges that centralized authorization engines like Permify work well for access control decisions, but highlights a scalability problem: efficiently determining which resources a user can access across large datasets without resorting to inefficient database queries with massive IN clauses.
The commenter acknowledges that centralized authorization engines like Permify work well for access control decisions, but highlights a scalability problem: efficiently determining which resources a user can access across large datasets without resorting to inefficient database queries with massive IN clauses.
4 points
· stangineer
· comments
Discussion Summary
No comments available.
No comments available.
More Stories (34)
252 points
· pumbaa
· comments
184 points
· johnbarron
· comments
231 points
· geerlingguy
· comments
297 points
· supermdguy
· comments
256 points
· loteck
· comments
62 points
· tzury
· comments
53 points
· andsoitis
· comments
41 points
· Cider9986
· comments
Ask HN: Who wants to be fired? (May 2026)
62 points
· evo_9
· comments
Ask HN: Who wants to be hired? (May 2026)
129 points
· whoishiring
· comments
33 points
· speckx
· comments
42 points
· Geekette
· comments
42 points
· throawayonthe
· comments
32 points
· robtherobber
· comments
82 points
· edent
· comments
206 points
· carlsborg
· comments
30 points
· XzetaU8
· comments
25 points
· jjmarr
· comments
Ask HN: Anyone feel like they're just opting out of tech these days?
22 points
· moomoo11
· comments
57 points
· ankitg12
· comments
223 points
· kodbraker
· comments
17 points
· crescit_eundo
· comments
203 points
· delichon
· comments
115 points
· deater
· comments
20 points
· greyface-
· comments
17 points
· cdrnsf
· comments
Ask HN: Who is hiring? (May 2026)
268 points
· whoishiring
· comments
107 points
· andsoitis
· comments
55 points
· bookofjoe
· comments
12 points
· apurvamehta
· comments
12 points
· theobeers
· comments
61 points
· beardyw
· comments
20 points
· PLenz
· comments
50 points
· moosia
· comments
Created by Zipper Data Co.
· 2026-05-02 12:00 UTC
· Unsubscribe
Get digests like this delivered to your inbox every morning.
Subscribe Free