Top 29 stories from Hacker News. Top 10 include comment highlights. Compiled at 20:10 UTC.
1097 points by signa11 · 406 comments
What HN said:
bottlepalm: I've hit this point with AI where it's not a simple process, but a long drawn out back and forth. I'll use AI to design the implementation of a medium sized, cross cutting feature. Review all the details, maybe iterate on just that. Then implement with Claude 4.
hrideshmg: As a junior, i do actually enjoy going back and forth with the AI discussing different ways to implement something and exploring alternatives. More often than not, I'd have an architectural idea that I'm not that confident in.
TACIXAT: This article doesn't address writing code with AI, just code review. My issue with agentic coding is that I make numerous micro-architectural decisions while programming. I almost never have a full spec up front and develop one as I consider what I am writing.
justinlivi: I find myself spending on average more time in LLM review/resolution loops than it would take for me to write the code by hand. Partially because once I'm in the flow I write very very quickly and the code pours out sometimes faster than I can write.
536 points by thm · 249 comments
What HN said:
solenoid0937: These - especially Polymarket - should be illegal globally, as they incentivize people with power to manipulate the real world in horribly destructive ways to win a bet. I would not be surprised if people are murdered at some point to reap the payout of some related bet.
throwawa1: When I see people making money on Iran attacks, and murder of heads of state - it shows clearly something is deeply wrong with Polymarket. Its a level worse than Vegas or Indian casinos. A literal ticket to hell. I'm all for banning these evil sites.
linuxhansl: Good. Just naming things differently does not work in other countries. If it quacks like a duck, swims like a duck, and looks like a duck, then it probably is a duck.
everdrive: I don't usually see advertisements, but I was in a position recently to see a real-life television stream, and I was quite surprised to see them run an advertisement for Kalshi. I was pretty surprised that something like this would be advertised to normal people.
434 points by vrganj · 168 comments
What HN said:
mcv: Finally! The entire country has been clamouring for this for weeks, and the government has been completely silent about it. A couple of weeks ago, the entire parliament (with only a single party dissenting) voted for a motion to end the contract with Solvinity, but the government...
bilekas: > "The politicization of this process has overshadowed the clear and important benefits this transaction would have brought to Solvinity's customers and Dutch citizens." That is unbelievably rich. It's politicians job to protect the privacy and interests of its citizens.
madbo1: This is exactly why privacy by architecture matters more than privacy by policy. The Netherlands trusted a policy ("Solvinity can't access the data") but the architecture allowed it anyway.
wildekek: As a Dutch citizen, I don't understand why we can't self-host an open source identity solution for 20M users with 30K requests an hour. How hard can it be?
349 points by subract · 65 comments
A visual explanation of how threshold secret sharing lets any k of n shares recover a secret.
What HN said:
946789987649: My masters thesis was on this! I created an app where you can store your data across all the common data storage providers (dropbox, google drive, onedrive, etc.) and used the secret sharing to aid with the encryption.
kristianc: Reminds me of this classic war story: https://medium.com/@amanat361/shamir-secret-sharing-the-stor...
c0l0: We use this technique in our team to distribute passphrases for our secondary secret stores (that contain instructions on how to access our primary secret stores) in a "democratically secure and safe" manner. https://packages.debian.
danborn26: Nice breakdown. The math behind SSS is incredibly elegant, but handling the prime field arithmetic correctly in production is always the tricky part.
280 points by dynip · 114 comments
60-second updates. RFC 2136 TSIG. BYOD. DNSSEC. For homelabs, edge routers, and infrastructure teams.
What HN said:
dynip: I'm Daniel, network engineer in Sweden. Built DynIP because every DDNS service I tried was designed around 2010-era networks: proprietary HTTP-only update protocols, poor IPv6, no DNSSEC, little support for actuallymodern devices.
basilikum: Pitch sounds really good. I don't have the time to try it out right now. However had I not read your comment pitching it here, I'd have closed the tab on the landing page immediately. Sorry to be so direct, but it just looks like any vibe sloped page out there.
jmusall: Refreshing to see competition entering this space. However, if you want to self-host, not caring for reliability or ease of use: bind9 supports RFC 2136 DNS UPDATE and DNSSEC, too (haven't figured that out yet, though).
hbogert: Bonus points for rfc 2136, works easily with external-dns. I've been using k8s+external-dns on-prem with a selfhosted minimal BIND server on a public host for years now.
244 points by croes · 213 comments
In this article, I try to understand why coding agents can be infuriating to use. I think the problem is their conversational UX: they behave enough like helpful colleagues to trigger our social instincts, but they don
What HN said:
zeumo: My take on the issue is that for most use cases where AI is pushed to the general public, a conversational chatbot is not the right tool, and the experience is bound to be frustrating. Remember when Copilot was basically a super-smart version of Intellisense? It was awesome.
xiaod: The compile-time vs runtime safety tradeoff is worth calling out. For infrastructure tooling where correctness matters more than iteration speed, the upfront cost pays dividends in reduced production incidents.
RandomBK: I've found swearing at a model to be quite effective in getting it to rethink and correct its mistakes. This seems to apply across Codex, Claude, Qwen, and Gemma/Gemini. I don't know if the model is picking up on a "need to lock in and be more rigorous" signal, or if the model pr...
em-bee: behaving like a human is not the problem. behaving unpredictably is. not doing what i expect, or rather not being able to define what i can expect is what's bothering me.
225 points by onemoresoop · 122 comments
What HN said:
arjie: It will be interesting to see what happens. One thing I really like about the US federal system is how each jurisdiction applies massive economic interventions. We get to run many experiments and see what happens.
e63f67dd-065b: The original impetus was more about banning robotaxis in Boston/MA than it is about the actual bargaining, from what I've heard. Just as the teamsters tried to ban cars to protect horse carriage drivers (that's what teamsters were, that's why they're called teamsters), they're ba...
devindotcom: Good for them. These companies appear exploitative and rent-seeking far beyond what the infrastructure they provide suggests is reasonable. If you're interested, next time you take a car, ask the driver what their end is - you may be surprised how little of the fare they actually...
nickvec: It's about time -- I hope rideshare drivers unionize in California as well. A UC Berkeley Labor Center study (2024) found rideshare drivers take home $7.12 per hour in median net hourly earnings before tips, which is less than half of California's $16 minimum wage at the time.
175 points by GodelNumbering · 193 comments
Outsourcing plus LocalAI will soon become more economical vs Frontier labs
What HN said:
jyounker: The problem with outsourcing, as opposed to remote developers, is that it takes a really good manager and tech lead to make it work. My experience is that you have to write extremely detailed design documents and work specifications in order to get effective results.
Chyzwar: When discussing LLM pricing, people are missing the plot. The subscription token price is 10x-40x cheaper than API pricing. Your 90$ Claude subscriptions give you close to $1000 to $4000 in equivalent API token pricing.
CuriouslyC: The future of American frontier AI isn't API calls, it's you taking your task to OAI/Anthropic like a consultant/external entity, then getting a product or whatever back, without ever seeing a large volume of intermediate work.
treis: I think this misses the forest for the trees. Working with ChatGPT is eerily similar to working with offshore Indian devs back in my enterprise days. Productive if guided explicitly but if let run wild there's lots of WTF moments.
162 points by aghuang · 161 comments
Dropbox CEO Drew Houston, who started the cloud storage company when he was 24, plans to step down and assume the role of executive chairman.
What HN said:
joshmn: If there are any Dropboxers here (drew—I emailed you a few weeks ago, but I imagine you're busy): I went to prison for 18 months, my digital and physical life was stolen from me: https://news.ycombinator.
gigatree: Board finally realized people can just do this themselves with FTP/SVN/rsync and curlftpfs
bhouston: Dropbox's stock has been stuck at around $6B valuation for years with flat growth and income around $2.5B per year. It is just stuck. Box.com, which is quite similar, is not that different. Around $3B and $1.2B in income. Similar valuation.
browningstreet: Having just rsync'd 100s of GBs back down from B2 and not sure where to put it, and having lots and lots of business documents and video files to share with collaborators, I'm surprised how few competitors there are in the Dropbox space.
145 points by juxtapose · 113 comments
Transformer-based large language models are increasingly used for long-horizon tasks; however, their attention mechanism scales poorly with context length. To handle this, we study a sleep-like consolidation mechanism in which a model periodically converts recent context into per...
What HN said:
pcrh: I can't pretend to understand how LLMs work, but I can be sure that anthropomorphizing their functions is not helpful to an objective debate over their abilities. Does a motor vehicle get "sleep" when it is serviced? When I reboot a computer, is that equivalent to a nap?
thunderbird120: The idea of periodically stopping to write blocks of recent context into a fast-weight state is interesting, but I think it liked it better when E2E-TTT[1] did it. It's a more flexible and elegant continuous learning approach.
bmc7505: This topic recently came up at the FLANN workshop [1], and seems to periodically be rediscovered [2,3,4] in different contexts. While some have speculated about the biological role it plays (e.g.
swyx: related preprint from the letta team https://arxiv.org/abs/2504.13171 Scaling test-time compute has emerged as a key ingredient for enabling large language models (LLMs) to solve difficult problems, but comes with high latency and inference cost.
AI/ML
Business/Tech
Other
Programming
Science
Stories and comments sourced from Hacker News public API. Not affiliated with Y Combinator or Hacker News.