HN user
https://www.mwaterman.dev/
It seems that they're trying to push up-market, or at least they were.
Given the extremely competitive releases of GLM 5.2 and DeepSeek V4 (both pro and flash), I don't think there'll be appetite for it.
It's a bit disheartening to see no comparison to other models here - and I'm not sure this pushes the curve anywhere. 3.6 flash is more expensive than GLM 5.2 - but seemingly worse, although this post is really light (lite?) on details.
It seemed for a time that Google had finally gotten the ball rolling, but I'm doubting that more and more as time passes. We'll see what happens with 3.5 pro I suppose.
There is no joy of discovery of new music that you haven't heard. There is no connection to other humans through music.
Hard disagree here.
For new music, Discover Weekly is great, if you take some time to engage with it on a routine basis. Even better, if you have an artist/genre you already like, the Fans Also Like or Discovered On will link you to other artists and playlists. Super easy to go down rabbit holes of new artists and playlists.
As far as connecting with others, I do like the spotify DMs (in-app share), the friend activity tab, and particularly the share attribution. When you share a song via link (url with ?si=), you'll permanently be linked to it. For a number of my favorite songs, I see "From {friend}" at the bottom while listening. Makes me feel super connected to friends I've bonded with over music.
Yes - this is certainly true. I would have to concede that I could not speak to how my thesis would map onto a truly colossal codebase.
That said - I would (again, maybe naively) suppose it's not hugely different - much of the work I do occurs in code where many people have and will work on it, and where the size of the codebase dwarfs model context windows.
In that case, I feel the same - current frontier models, when properly oriented to a task, with some assist on the big-picture thinking - are more than capable of generating good code that can slot into big codebases with many moving pieces. Of course, I'd have to point to other people's work to defend this, but I think that's still pretty reasonable especially against the declared "LLMs are worse than useless for generating code".
Obviously I'm just one guy, but I maintain what I said w/ typing happily at 120+wpm. I think I could break 150 if I switched to colemak or dvorak, which I've been considering.
I don't think my job was ever to type fast, nor do I make any claim that I'm ontologically better than someone writing code by hand - but for what I need to do, I'm way faster now.
You might think he's an AI shill, but I was pretty compelled by simonw's post and idea - "your job is to deliver code that works" [1]. I think I can deliver more code that I can prove to work, faster now. The productivity is nice, but as I've said on other parts of the thread, it's also just fun to spend more time thinking, less time hitting the semicolon on my keyboard?
[1] https://simonwillison.net/2025/Dec/18/code-proven-to-work/
Edit: should have used [1] in the first place instead of an asterisk.
Yes - I think that's fair. I'm happy to concede the term coding - I still enjoy the process of telling my computer to build something I've envisioned. It's now much faster, and I get to do more of the big-picture thinking and ideation, which I thoroughly enjoy.
I guess I'm of two minds on it.
Sometimes, I will review every line, test the front-end in a staging environment, verify the backend contract, et cetera. Over time, though, I realized that many of these reviews just didn't result in any necessary changes. The current model (with guidance/claude.md/etc) was able to one-shot the task.
Not to overly personify, but imagine how you might treat a junior colleague. You start by reviewing everything they do with a microscope, later you review the broad-strokes, and eventually, for low-stakes or well-scoped tasks, you just play with the demo and the ticket and approve it.
Otherwise it's not materially different than a pre-AI world - you've got sample I/O, test cases, hand-review, look at the application on different screen sizes, contrive some edge cases, test against a spec if there is one - et cetera.
I love coding and I always have - arguably moreso now, if you can still call it coding.
For me, it is not about the syntax nor the mechanics of typing though. My enjoyment comes from thinking about problems and breaking them down, or thinking about what architecture would best serve them. I guess I'm meant to be a systems design guy, so I'm lucky that AI-coding fits this well (AI models have limited context windows, relative to the size of codebases, so doing the big picture thinking is still important, and fun for me).
Generally end-user applications, depending on how you classify internal tools. I'm generally very happy with the output of Opus 4.X with a moderately structured CLAUDE.md and some investment in detection/avoidance of anti-patterns (Next.js+ts). I imagine that the bitter lesson is true here, and these heuristic guidelines will become increasingly unnecessary w/ smarter models.
Library-type work has mostly been side/toy projects, although fwiw, with a standard/spec on hand (CommonMark for example), I'm also happy w/ the output. It's often possible to "close the loop" and have the coding agent autonomously iterate until the standard is adhered to.
I can only imagine that people who say things like:
If you use AI for anything else, and in particularly if you use it to generate code, you're wasting your time.
Have not used frontier models in at least a year.
It is nearly inconceivable to me that I would ever go back to writing code by hand, in any context. Even if no new model was ever released, the combination of GLM 5.2 and DeepSeek V4 flash is more than sufficient to do real work. It's not hard to imagine that a distillation of Mythos/5.6S or whatever is a couple generations away will push me even further in this direction.
+1 for the claude icon fire bars
I would add a +1 for testing w/ Word - the official Office suite runs some validation where only Word will show a "broken file" popup, even when nothing else does.
In our case, clients use only real Word, so any machine-generated/mutated files (excel/ppt as well) need a pass through the real office executable.
Interesting - I'll be sure to benchmark it at some point. We've found the best results come from blending providers depending on the task anyways.
Thanks for the quick response - and always happy to see more competition in the space. Best of luck with future features!
Unclear what difference exists against Firecrawl - their team has been shipping great features extremely quickly lately, and their core offerings have become really good.
I am interested in KnifeGeek though - looking for a good OTF (ultratech?)
Nor more than a mention of Query Hints, which had some interesting discussion under a similarly-titled submission.
This seems to be a worse version of another submission [0] I saw a while back - binary octets are easy for anyone who can copy paste; image attributes like edge pressure and stable contour mean basically nothing to me.
No one takes g2 and product hunt seriously right?
I think Mythos is rumored to be ~10T parameters, so in this case I think the answer is yes, although I'm sure MoE, looped models, etc play a role in the improvements as well.
AlphaEvolve’s procedure found an algorithm to multiply 4x4 complex-valued matrices using 48 scalar multiplications, improving upon Strassen’s 1969 algorithm that was previously known as the best in this setting. This finding demonstrates a significant advance over our previous work, AlphaTensor, which specialized in matrix multiplication algorithms, and for 4x4 matrices, only found improvements for binary arithmetic.
And in 20% of cases, AlphaEvolve improved the previously best known solutions, making progress on the corresponding open problems.
https://deepmind.google/blog/alphaevolve-a-gemini-powered-co...
Very cool concept, although I think the leaderboard is more about network latency than anything else...
Although I'm anti AI-image-gen (wouldn't there be CC0 images of these species?) this is really quite a charming little project.
I feel like I don't hear many bird calls in the city where I live, but maybe I'll have to set one up for myself to find out!
Not all the inputs, but Ben Felix’s company (makes videos on this topic) has a rent vs buy calculator, mainly focused on investing the cost difference for mortgage vs renting: https://research-tools.pwlcapital.com/research/rent-vs-buy
This is really getting ridiculous - although people sometimes dismiss the "missing" status page because it includes copilot, it's worth noting that pull requests (95.5%) are even lower availability than copilot (96.4%).
How am I expected to comment "LGTM" if I can't even get to the PR?
Although this affects a small number of people / kids per year - this is a great development. There's a related article^ which had a very moving account from a mom whose son has received this treatment.
^: https://www.npr.org/2026/04/23/nx-s1-5795526/deafness-gene-t...
Almost refreshing to see some good old-fashioned schizoposting, which usually doesn’t end up on HN.
Really interesting response to Google's Stitch - and seemingly a better alternative given some of the features shown in the video. If everything actually works at least close to how it's advertised, this'll be useful. I'm sure it's no Dieter Rams, but it wouldn't be a surprise if it's already better than many devs at design work.
Obviously the discussion here is mostly about writing code. In that domain, I’m always of two minds on this sort of thing. Although I think everyone would agree that material cognitive decline is bad, I also think we have to be precise with what that means.
During university, for an exam in a graduate databases course, I had to manually calculate the number of operations for a query, down to the ones place. We were given an E-R diagram, the schema, and the query. So we had to act as the query planner - build out the B+ tree, check what was most efficient, and do it.
This is by all means a pointless endeavor - no one has had to do this by hand in literally decades. It was also among the hardest cognitive tasks I've ever had to do. After being one of two people to complete the exam in the three allotted hours, I sat outside the lecture hall on a bench for a little while because I though I might faint if I went any further.
I’m beginning to feel the same about writing code by hand. If I can design systems that are useful, performant, and largely maintainable, but the code is written by an LLM, is this harmful? It feels that I spend more time thinking about what problems need to be solved and how best to solve them, instead of writing idiomatic typescript. It’d be hard to convince me that’s a bad thing.
Some nice tips in here ([field: SerializeField] for example) - but as others note, it's not about modern code, as many of these features have been available for an embarrassingly long time. It's always felt to me that there's some fundamental friction between idiomatic C# and Unity.
I remember, after reading about new features C# 8.0, someone wrote that C# 9 would write all your code for you, and that C# 10 would just mail you a check every month. How the times have changed...
Die tests must be 100% accurate and follow the best known clinical procedures. Humor is not optimal.
I don’t think there’s a reliable system or API for doing so, unclear that arms race will ever favor the side of the detectors.
As far as how I / other people do it, there are some obvious styles that reek of LLMs, I think it’s chatgpt.
There’s a very common structure of “nice post, the X to Y is real. miscellaneous praise — blah blah blah. Also curious about how you asjkldfljaksd?"
From today:
This comment is almost certainly AI-generated: https://news.ycombinator.com/item?id=47658796
And I'm suspicious of this one too - https://news.ycombinator.com/item?id=47660070 - reads just a bit too glazebot-9000 to believe it's written by a person.