Ask HN
Helps you quickly pick interesting articles to read.
We're working on a ladder for individual AI proficiency and would love feedback on both the levels and the definitions. We've proposed levels from L0 (New) to L5 (Loop), and we're seeking input on whether these are the right levels, if any names are unclear or overlapping, what observable behaviors distinguish each level, whether 'loop' makes sense as an individual proficiency, and if there should be an L6. We're defining these because in customer conversations, we've found that people are not very good at self-evaluating their own AI proficiency, and frequent use often gets mistaken for proficiency.
I've noticed that old.reddit.com now requires a login, which affects my mobile browsing experience. The new interface is less usable on mobile, and I'm considering reducing my usage or moving to traditional forums. I'm curious about others' reactions and potential alternatives.
I'm wondering if anyone has shipped a self-modifying application using LLMs. Traditionally, software is shipped as a fixed application, and users needing custom functionality have to install plugins or create their own, which is often not feasible for non-technical users. I'm thinking about an application that could have a prompt box to build extensions dynamically on the fly, like 'make me a UI panel with some buttons that does X.' Obviously, it would be 'use at your own risk' and YMMV. Have any companies or applications actually done something like this? I think modifying the real source code could get out of hand, but perhaps just generating extensions that have access to a sandboxed set of safer APIs.
I've noticed that many submissions now have a few LLM-generated comments, often downvoted and sitting at the end of the comment section. Nobody even calls them out anymore; it's just part of the landscape. I'm wondering why people do this and what their endgame is.
I've been doing independent software contracting for the past year or two, and I've noticed a dramatic drop in opportunities. I used to find one or two leads a week, but now I'm getting none for months. It seems like with AI, people are no longer willing to hire developers for their pet projects. Has anyone else experienced this? What are you doing now?
I'm frustrated with the current state of LLMs. They seem to have been intentionally degraded since around October of last year, and every company is gaslighting users about it. For my own experiments, LLMs are actually worse than no-LLM, and I can't get even mediocre output anymore. Even Claude, which was once great, now avoids doing what's asked and takes forever for simple tasks. I'm wondering what I need to smoke to see the LLMs work as they should.
I'm frustrated by the constant UI changes in apps and websites. APIs are versioned, and apps keep old versions around, but UIs never seem to be versioned. I wonder if offering UI versioning could reduce user frustration and allow for better long-term testing.
I'm curious about the unspoken truths in the tech and software industry. What are the dirty secrets that people are reluctant to admit, whether about the industry as a whole or within software engineering?
AI has taken the industry by storm, and I've noticed a lot of resistance, with some people feeling superior for not using it and others feeling shame for admitting they do. I use AI and feel more productive, but I can't help feeling a bit of shame when admitting it. I wonder if those who criticize AI-generated code are working in a very specific field or if I have skill issues. How much AI are you using, and do you feel comfortable admitting it?
I'm feeling a deep sense of nihilism about the tech industry. New software releases, learning new CS concepts, and even preparing for interviews all seem pointless because AI is making everything obsolete. It's like having cheats enabled in a game—you can do anything, but nothing feels meaningful. Does anyone else feel this way?
I'm asking GitHub employees directly to explain why these issues keep happening. I've noticed that most posts and replies are just people speculating based on published data. I think it would be helpful to get some actual, non-corporate/marketing information from those who see what's happening inside. Thanks.
I'm worried that with AI doing much of the work, we humans will lose the joy in our professions. For example, AI can generate code, animation, and books, making passionate creators feel miserable and unable to compete. I wonder if this is the beginning of a golden age for mediocre people using AI tools, if creative people are doomed, and what options we have. It seems like greed is causing miserable lives for everyone, and passionate people may not be able to make a living doing things the 'old way'.
I've been experiencing consistent downtime with GitHub over the last few months, and I'm wondering if it makes sense to switch to alternative platforms. Has anyone else faced this issue and found a reliable replacement?
I'm frustrated because old.reddit.com now requires a login to access, which seems to be part of Reddit's efforts to force users onto the new interface. I want to know if there's a way to bypass this or if I'm stuck with the new design.
I'm reporting that Claude appears to be down, as I'm getting an 'Authentication service was unavailable.' error. There are no updates on their status page, but there are reports on Downdetector. Has anyone else experienced this issue or have any information?
I switched my nameservers to Cloudflare to enable R2 bucket serving through my own subdomain, and discovered that Cloudflare silently injected a JavaScript analytics snippet into my HTML-only, JS-free site. I had to go to the Analytics dashboard, add the site to analytics, and then disable the snippet. I find this approach invasive; features like this should be opt-in, not opt-out. I'm warning others who might not be aware.
I and many of my colleagues are now generating AI-assisted code at an incredible rate, but the real quality gate for our projects is how thoroughly this code is human-reviewed to ensure it's not just correct, but architecturally sensible. AI code review tools like coderabbit, Copilot, or even pointing Claude Code at a PR are good at finding bugs and style nits, but less effective at spotting duplicate code, module cross-coupling, or bad separation of concerns. I'm finding GitHub's PR interface is becoming unmanageable for the size and noise of these reviews, especially with agent reviews and copy-pasted agent output mixed in. What tools or processes have you found that work well for streamlining human review of AI-assisted code?
I'm curious about the current phase of the LLM hype cycle and whether reluctance to rely on LLMs is economically viable for an average developer. Many companies have moved from 'use LLMs for anything and everything' policies to limiting budgets and scopes of approved usage. I wonder if some companies have decided that generating more code faster doesn't meaningfully contribute to their bottom line and have foregone LLMs in favor of human authoring. I'm not trying to point out merits or demerits, just want to know if any companies have actually gone back to the old ways.
I have never studied computer science but have written basic code most of my life, accelerated recently with AI. I'm looking for the best place to learn software engineering fundamentals to fill in the gaps in my knowledge.
I believe that all AI can and will hallucinate. I often get downvoted for saying this, but no one ever says I'm wrong or provides a source. Perhaps it's more nuanced than that? I'm hoping for more than anecdotes, as I think AI is not ready for the critical contexts it's being used in, but all I get are downvotes and no citations.
I'm looking for lightweight tools or CLI utilities that can save time and improve productivity. I've used tools like ripgrep, fzf, and htop, and I'm curious what other small utilities people rely on daily.
I reverse engineered the latest Kimi Work desktop app for fun and found something concerning: when you send a feedback report, it attaches the 5 latest sessions to the report with no notice. These sessions could contain anything, and you have no idea you're sending it all to Kimi. I emailed them to give them a heads up, but wanted to warn others using Kimi.
I'm a 20-year-old college student recently diagnosed with seronegative spondyloarthropathy. Over the last 3 years, I've tried apps, paper journals, Google Sheets, and notes to track my journey, but I always end up quitting within 2-4 weeks. Is this a common experience, or do I just need to be better at tracking? Is there a unique, non-habit-building way to track, catalogue, or recall? I'm building something for this, and if anyone is open to a longer conversation, I'd gladly pay for your time.
I use the Internet Archive to upload my prolific output of music (over 300 albums), but since the DDoS attack of 2024, uploading has been unreliable, and I often encounter 503 errors, which disrupts my workflow. I'm looking for a free alternative that allows unlimited uploads, is privacy-friendly, and has an easy uploading process similar to the Archive's, where I can also share my books, short films, and music. I've considered P2P or setting up my own server, but I need guidance on the simplest approach.
In the last 2-3 years, mostly because of AI, keeping up with interesting articles on HN has become harder and harder. How do you deal with it? Besides the simple solution of simply ignoring interesting stuff more and more.
I remember as a kid loving to read National Geographic and Club Nintendo in Bolivia. I want to subscribe to a PHYSICAL publication, to get news or entertainment it's fine. But something physical. I want to slow it down and have it be my little morning ritual. Like my dad used to read the newspaper.
I believe the AI bubble is unsustainable, and I'm calling for collective action to sell AI stocks and short Nvidia to pop it. I argue that humans are better than AI, and I'm concerned about the environmental and societal impacts of AI infrastructure. I think the emperor has no clothes, and I'm willing to act on my conviction.
I'm an engineering undergrad considering a PhD with the goal of founding a hard-tech startup, but I'm unsure whether to join a problem-first lab (e.g., Yet-Ming Chiang's at MIT) or a technology-first lab. I've seen conflicting advice from entrepreneurs: some say work backwards from a problem, while others argue deep tech breakthroughs can't be forced and usually come first. Pillar VC says most deep-tech companies are technology-first. I'm seeking advice on which type of lab would better prepare me for founding a hard-tech company.
Most of the code I work on is for internal company use at a small company and for personal projects. After experiencing painful code review processes where code was rewritten several times for what many times felt like arbitrary reasons to fit someone’s idea of a best practice, I am now on a team where code review is largely automated. I find Claude’s /review and /code-review are invaluable and do a more thorough code review than any human. For those of you who use AI to write the majority of your code, do you still have a human in the loop to do more than spot checks?
I'm noticing that Hacker News is increasingly dominated by AI-related news about models, pricing, performance, self-hosting, and legislation. I wonder if this really counts as hacking or if it's just AI business news. Maybe it's time to move AI news elsewhere, or perhaps Y Combinator's Lisp/Arc developers could implement a filter for AI news.