nayroclade

If you include a word like innovating in quotes it typically implies that you're quoting it from the link. It can also signify irony, but in a context like HN where we're discussing a published article, it's often ambiguous.

As for Java, I'd agree that its pace of advance was pretty glacial during the Sun era, but from what I've seen has picked up considerably since the Oracle acquisition and Brian Goetz became architect.

And however bad Java is, it's nothing compared to JavaScript. It takes a decade just to add new a library function, and every new syntax proposal is DOA.

I didn't mean it as condescending. I meant it literally is cute: A neat idea that is quite cool in its execution.

Cute idea, but you're never gonna blow your token budget on output. Input tokens are the bottleneck, because the agent's ingesting swathes of skills, directory trees, code files, tool outputs, etc. The output is generally a few hundred lines of code and a bit of natural language explanation.

Did Anthropic run Claude in a loop and tell it "work on our pricing and usage strategy", or something?

As far as I can tell, they now have Free, Pro, Max x5, Max x20, Team, and Enterprise tiers. Plus session usage, weekly usage, extra usage (up to a spend limit (set by you) and/or a monthly cap (set by Anthropic)), and now usage bundles, which are extra usage but with a lower price and a fixed amount.

I don't know, but it really feels like there's an LLM churning away somewhere, endlessly tinkering on a PRICING.md file that is gradually accreting strategic slop.

> I don't know how long this pace will last. I suspect that bugs are reported faster than they are written, so we could in fact be purging a long backlog

Hopefully these same tools will also help catch security bugs at the point they're written. Maybe one day we'll reach a point where the discovery of new, live vulnerabilities is extremely rare?

Nah, blaming AI is too easy. It's more likely that Apple's design culture got rotted out under Alan Dye https://daringfireball.net/2025/12/bad_dye_job

Now that Dye is gone, I still hold out hope that Apple will change direction and start fixing their UI. But that fact that it got this bad in the first place implies things are seriously broken at a senior leadership level.

Or just consider the “asshole dinner guest” trope that appears in so many TV shows and movies. They will always be talking too loudly and gesticulating/pointing with their cutlery.

I’m pretty sure I’ve used cutlery pretty close to this in some hipster restaurants.

There are plenty of web devs who care about performance and engineering quality. But caring about such things when you work on something like a news site is impossible: These sites make their money through user tracking, and it's literally your job to stuff in as many 3rd-party trackers as management tells you to. Any dev who says no on the basis that it'll slow the site down will get fired as quickly as a chef who get a shift job in McDonalds and tries to argue for better cuisine.

The WP reportedly lost $100m in 2024. So one the one hand, you might understand Bezos wanting things to change. On the other hand, Blue Origin reportedly loses multiple billions of $ per year, and has done for decades, which Bezos pumps in without insisting on massive cuts or layoffs.

Annex 1 of the study lists the test results for individual brands/models. There are several, including Airpods, that get an all-green evaluation score.

I suspect AGENTS.md files will prove to be a short-lived relic of an era when we had to treat coding agents like junior devs, who often need explicit instructions and guardrails about testing, architecture, repo structure, etc. But when agents have the equivalent (or better) judgement ability as a senior engineer, they can make their own calls about these aspects, and trying to "program" their behaviour via an AGENTS.md file becomes as unhelpful as one engineer trying to micro-manage another's approach to solving a problem.

There's a big difference between being in a field versus working in a field, from dawn to dusk, every day, regardless of the weather or sickness, in order to produce just enough food to feed you and your family, knowing that a single failed harvest (due to conditions like weather and pests that you have no control over) will leave you starving, watching helplessly as your children, spouse, friends and neighbours slowly weaken and die, knowing that even if you survive, you will face the same thing again the next year, and the next, for the rest of your lives, which will likely be short, due to the constant, exhausting labour, frequent bouts of malnutrition, and nonexistent medical care.

I expect its synergy with Tailwind. Shadcn/ui uses Tailwind for styling components, and AIs love Tailwind, so it makes sense they'd adopt a component library that uses it.

And it's definitely a real effect. The npm weekly download stats for shadcn/ui have exploded since December: https://www.npmjs.com/package/shadcn

Firstly, an auto paint shop is not the same as an auto manufacturing plant.

Secondly, it says you can't permit a new auto paint shop in CA, but it specifically mentions the Bay Area AQMD as the reason. But, as its name implies, the Bay Area AQMD only regulates within the San Francisco Bay Area. It is only one of 35 air districts in California: https://en.wikipedia.org/wiki/List_of_California_air_distric...

So, it is impossible to permit a new auto paint shop in all of these districts, or just the bay area? Because those are very different. It also labels starting a new paint shop as "impossible", but then says it's "nearly impossible". So is it actually impossible, or just nearly impossible?

> Inexplicably, until 2025, browsers stuck with the naive greedy algorithm, subjecting generations of web users to ugly typography.

> WebKit devs, you are awesome for shipping this feature ahead of everyone else...

Um, no? Chrome shipped this feature in 2023: https://developer.chrome.com/blog/css-text-wrap-pretty

Safari isn't early shipping this, they're late. Though not as late as Firefox, admittedly.

Is the approach fundamentally limited to smaller models? Or could you theoretically train a model as powerful as the largest models, but much faster?

Hooks were introduced in 2019. so, seven years ago.

Arguments against proactive MRI scanning always seem to have a whiff of status quo bias to them. Yes, right now MRIs are expensive and false positives are common, but if regular scans were widespread, it's likely this result in innovations that would drive down costs, improve accuracy, as well as producing a much larger corpus of data with which to guide diagnosis and reduce false positives.

To use a software analogy, if your downtime detection system kept producing false negatives, would your solution to be just turn it off? You'd get some better night's sleep, but you'd pay for it when the system really went down and you had no idea.

When you built this exact product, how long did it take you to reach 80% compatibility?

This isn't really true though. Pre-training for coding models is just a mass of scraped source-code, but post-training is more than simply generating compiling code. It includes extensive reinforcement learning of curated software-engineering tasks that are designed to teach what high quality code looks like, and to improve abilities like debugging, refactoring, tool use, etc.

NATO could certainly rollover the Russian army in a conventional war, but that was just as true before the Ukrainian war. The idea that Russia is/was a serious threat is a convenient fiction: It helps maintain Russia's image as a superpower, and it provides a justification for the existence of NATO and the associated military industrial complex that supports it.

What is true however, is that Russia does possess a huge arsenal of nuclear and other weapons:

https://en.wikipedia.org/wiki/Russia_and_weapons_of_mass_des...

Despite Putin's posturing, Russia's never going to risk deploying them in a conflict with Ukraine. But in an actual war between NATO/Europe and Russia, with the regime facing an existential threat, then there's a very good chance they would. But even before it got to that point, the nature of the conflict itself would make nuclear escalation very likely. Both sides would be firing huge numbers of missiles, attempting to gain air superiority by wiping out the other's own missile launchers, radar bases, etc. With that many missiles flying, and stressed people and automated systems making split-second decisions, it's very likely that an error or miscalculation would result in an accidental nuclear strike, at which point it would be impossible to put the genie back in the bottle.

Telling people to only run OpenClaw in a full isolated sandbox kind of misses the point. It's a bit like saying, "gambling fine so long as you only use Monopoly money". The think that makes OpenClaw useful to people is precisely that it's _not_ sandboxed, and has access to your email, calendar, messages, etc. The moment you remove that access, it becomes safe, but also useless.

There are some ideas to produce something like parameterised querying for LLMs, such as DeepMind's CaMeL: https://simonwillison.net/2025/Apr/11/camel/

If AI can code, why do you think it cannot handle building, testing, debugging, securing, monitoring, supporting, incident handling, and so on?

Consider incident handling. What if your AI sets up monitoring that detects errors or outages, wakes up an agent, gives it the problem context, then sets it to work so it can debug the issue, produce a fix, then deploy it? You now have an end-to-end system that works 24/7. Many issues will probably be resolved before you've even noticed them.

If your response is, AIs won't ever be smart or capable enough to do this as well as humans, how has that same prediction worked out for coding?

The next generation of AI "coding" tools will essentially be SaaS companies in a box. Agents will code the app, but they'll also test it, debug it, support it, etc. And this will happen in months, not years.

The models they tested are already way behind the current state-of-the-art. Would be interesting to see if their results hold up when repeated with the latest frontier models.

The meta-analysis cited in the article is from 2024 and specifically mentions the Shaffer et al. 2014 review cited by Wikipedia as being low quality:

> Some of the available reviews, owing to the limited number of trials and methodological biases, were of low quality (Anglin et al., 2013; Cheng et al., 2020; Li et al., 2014; Shaffer et al., 2014).

https://pmc.ncbi.nlm.nih.gov/articles/PMC11650176/

I always found the "Amazon 4-Star" name funny. Presumably when it was first pitched internally it was called "Amazon 5-Star", then they realised that meant they basically couldn't sell anything, since nothing popular gets a full 5 stars. So they changed it to "4-Star", which just sounds awkward, and lacks the suggestion of top-quality that "5-Star" would. Instead, it's like the "Amazon Not-too-bad" store. I was amazed that they actually went ahead with it.

Slop Tuesday

They have an answer to that in their FAQs: https://proposal.eu-inc.org/FAQ-Glossary-14d076fd79c581d18e6...

> While both aim to facilitate cross-border operations within the EU, EU-Inc addresses some of the limitations of SE:

> - No minimum capital requirement: Unlike SE, which has a minimum capital requirement of €120,000, EU-Inc has no minimum capital requirement, making it more accessible for startups.

> - Simplified governance structure: EU-Inc offers a more streamlined governance structure compared to SE, reducing administrative burden and promoting flexibility.

> - Digital ecosystem: EU-Inc is supported by a robust online ecosystem, including a digital registry and dashboard, for efficient management and compliance, which is lacking for SE.