KolibriFly

That's why a hybrid approach is needed. The agent shouldn't be making up dimensions based on an image. It should use OCR to extract the size table from the datasheet, feed it into a parametric table, and only then map it onto the base enclosure template.

Comparing it to Leetcode is completely off. On Leetcode, you're running an isolated algorithm that doesn't need any access to internal repos or staging environments. But a CI/CD sandbox that reviews PRs by definition has to have serious permissions, or it won't be able to build anything.

Even if you leave intent aside, the effect is the same: it teaches researchers that funding is conditional on staying within an invisible and shifting political boundary

And scientists are often exactly the kind of people who will try to keep going anyway

A research system can adapt to lower funding if the rules are stable. What it can't adapt to is grants being frozen, staff disappearing mid-project, forbidden vocabulary changing

A degree was never supposed to make someone instantly productive on day one. It was supposed to be paired with junior roles, mentoring, code review

I had a similar experience with unfamiliar frameworks. The less you understand the domain, the easier it is to accept clean-looking but wrong code

Good test actually : try explaining the solution without any hints. If you can't walk through why the code is structured that way, then the review was superficial

That's probably one of the most sustainable approaches. Queries like show me where to look for the problem often turn out to be more useful than just fix it for me, because they leave the most valuable part of the work to the engineer

Hardware scarcity and cloud incentives are real, but they're also part of the same broader trend: more dependency on centralized providers, less control for the buyer

Just don't design the game so that, when the business model stops working, every paid copy becomes a brick

This is exactly what is happening right now. Models are becoming more efficient but at the same time users are starting to tackle tasks they previously didn’t even try to automate.

But Jevons paradox explains the increase in consumption, but it does not necessarily answer the question of business profitability

Well, the history of cloud computing shows that infrastructure usually becomes cheaper over time. But it’s still unclear whether this rule applies to reasoning models.

It seems to me that both sides are starting to drift into extremes. Some promised the replacement of half of office workers, while others are now saying that AI doesn’t create any value at all. The reality is somewhere in between

Sadly not quite one per wheel, though that would make the title even better

Self-balancing bicycles, sure. Self-driving bicycles that navigate city streets safely are a much larger problem.

I think that's basically right. The "neurons" framing feels a bit loose here; it's more like a very compact hand-designed controller expressed in neural-network-ish terms

The useful insight is not "compare desired state to actual state"; it's deciding which state to control

Yeah but at least early industrial machines consistently outperformed humans on a specific task. With LLMs it's still a weird mix of brilliant moments and completely unreliable behavior

It's strange how much attention novel psychedelic treatments get compared with older, less glamorous interventions that already help a lot of people

I read it less as "this only helps veterans" and more as "veterans are the group this particular research and funding path is centered on"

It feels promising, but also exactly the kind of treatment that should move through careful clinical trials

This is exactly why Proton feels like the pragmatic path. Native ports are nice in theory, but PC games are rarely just one clean executable anymore

This is such a neat historical parallel

Primary energy comparisons can make fossil fuels look more "irreplaceable" than they are, because so much of the input energy is lost as heat before it becomes useful work

Factory games are weirdly good at teaching the "shape" of these systems

That must be an incredible place to grow up around

The odor point is interesting. I think a lot of people mentally picture refineries as visibly dirty and smelly by default, but a plant near dense urban/residential areas probably has very strong incentives to be almost boringly well-contained

What always strikes me about refineries is how "simple" some of the core ideas are in isolation

This is probably a combination of ergonomics and expectations

The interesting part isn't whether Apple got the design right or wrong, it's that most of us never even consider altering the tools we use every day. We just adapt ourselves instead.