HNHacker News
TopNewBestAskShowJobs

elxr

407 karma · joined November 25, 2021

submissionscomments
elxr··on New iMac with M6
The studio display is priced like a ultra high end monitor but gets beat by $800-900 monitors all day. It's small, it has bottom of the barrel 60Hz, and it's only saving grace (the resolution plus the build quality) is made way less exciting by the fact it's not a modern OLED with amazing HDR.

Too many great options today to settle that overpriced undersized thing.

elxr··on GPT-6.1 Terra
Run your own evals then if you have this much of a need to "correctly" choose. That'll make it clear.

I'll just stick to mostly Sol on low/medium/high and start experimenting with Terra tomorrow onward to see what it does noticeably better than Luna/Deepseek flash.

Any pareto frontier model, especially from the leading labs is guarnateed to be good enough for whatever your expectations are at this point. A quick few day test should be enough for anyone to make their own conclusions.

elxr··on CSS-Tricks has a new owner
CSS tricks is a shining testament to consistency and not falling for trend-chasing. I love how laser focused they are on covering purely CSS and making it fun to keep up with CSS news. Sounds like they're on a good path right now, and he's convinced in his choice to take full control.

Super happy the site.

elxr··on GPT-6.1 Terra
The 4-tier naming is honestly as clear as you can get. I have no idea what people are so confused about.

Terra and Luna are dirt cheap and good enough for 50% (or more) of the typical junior dev's workload, and great for routine work. Sol and Astra have so much written about them at this point that it should be obvious what those are good at. The large/gigantic models for things that need full intelligence and slightly more branching thinking.

elxr··on GPT-6.1 Terra
If you daily all these models (and have a modest budget, not some >$400 corporate allowance), then it's pretty obvious that Luna is small and fast, and Sol 6.1 is the big one for jobs that need significantly above average intelligence, comparatively to a human engineer. I use both constantly. Astra is really not relevant for the average dev except to do some moonshot task with a vague verification loop.

Sure Sol 6.1 on low/medium can be considered "cheap" considering its abilities, but it's clearly not meant for the same type of tasks as Luna. You just grow an intuition for when something is solvable with a Luna-sized model. I'm guessing Terra is good for the tasks that you'd typically use Sol 6.1-low for, or for Luna type tasks (or other not-novel problem) where you want to reduce the risks of suboptimal decision making. It's exactly what I'd go with for feature additions where the correct solution is already mostly designed/specced-out.

elxr··on Oldest GUI application that still runs on Windows 11
Functionality isn't hidden and buttons looked like buttons (instead of the modern trend of having unstyled text or tiny icons having a 50-50 chance of being a button).
elxr··on The Goose Programming Language
That's a great idea. I wonder if any major library (or language) has attempted a version their site in that style.

I'd love to experiment with this myself.

elxr··on The Goose Programming Language
> The documentation is perfectly comprehensible

For a programming language? It's subpar and you know it.

See this for a language docs site by a small-ish team, but a significant effort was put into making it nice to read: https://tour.gleam.run/everything/

elxr··on Show HN: UI micro interactions prompts for AI
A good chair's expensive? I don't agree.

I'm not talking about upper luxury products. I'm talking about good, maintainable business code or just something as basic as a polished, full-featured dashboard that contains no "silly" bugs.

That kind of thing is still not very cheap in the context an average person's typical expenses.

The chatGPT webapp is as example of something built to a budget (despite having a team with access to cutting edge AI); and it just doesn't feel good to use or to navigate around. Contrast that to Twitter's front-end. That took millions of dollars in designer/SWE salary to build and you can just tell how much better it is than the average social media platform, or any new webapp by a less experienced team.

Human in the loop prompting/steering/testing time is also not free, even if a $20 claude/chatgpt sub can theoretically build anything to an acceptable level. 2 days worth of a humans's time is still ~100 bucks. Very few harnesses can run fully autonomously and still end up with usable software unless you're someone who specializes in working with these things.

Production level code is really not as cheap as you imagine it to be.

elxr··on We should be more ambitious with the software we build
100% agreed.

There's so many youtube channels getting hundreds of thousands of views on EVERY new major model release just doing essentially zero effort one-shots, and people lap it up as if building things that already exist is somehow the best their imagination allows for.

I do love being able to test a model for 3+ weeks (>1 month is too long personally for my budget) and actually learning how to build scalable software and actually develop your "engineering" skills.

elxr··on Dessa: Conversation Model trained only on speech
Authentic to what? That annoying clinger who loves interrupting you when you're trying to speak in turns?

Is having a duplex model constantly cutting in with uninteresting quips something people are looking for?

elxr··on Kimi K3.1 Pro and K3.1 Pro Highspeed
This'll end up on opencode go soon enough. $40/month on there gives you a bit more than just "light use"

Don't use frontier models like this for problems that can be solved with flash models, and you'll see API pricing isn't that terrible. Still cheaper than anything with remotely this level of capability 2.5 months ago.

elxr··on OpenShell – A safe, private runtime for fleets of autonomous AI agents
If that isn't a sign to stop using anthropic then I don't know what is.
elxr··on Show HN: UI micro interactions prompts for AI
> the value of code is very low

Good front-end code is still expensive. Do you think ChatGPT/Anthropic/Meta/etc. can claim industry leading front-ends on their webapps or desktop apps? Why have they not reached parity with all their smaller competitors if it's all just code.

There's also new serverless platforms constantly being funded nowadays; if code is so cheap why doesn't everyone just whip an LLM to go build their own version? Only bad code is low value.

But that aside, $30 per month is questionable.

elxr··on Show HN: UI micro interactions prompts for AI
All those NFTs are literally just small images though. You're guaranteed to get the exact same image whether or not you "own" the NFT.

Running claude yourself to replicate UI is an actual time investment and takes some setup.

elxr··on Show HN: UI micro interactions prompts for AI
Nice examples, definitely lots I'll be copying. I do wish there was an option to set the number grid columns on the page though: a sort of zoom-out feature with added column counts.

I feel having a lot more items visible without scrolling would be perfect for a site like this. Especially with the focus being entirely on the rendered components (and not typical UI library docs).

elxr··on Novo Nordisk, Eli Lilly Report GLP-1 Signals of Slowed Biological Aging at ARDD
Not my place to talk about.
elxr··on Novo Nordisk, Eli Lilly Report GLP-1 Signals of Slowed Biological Aging at ARDD
> i’m also amused you couldn’t help yourself from saying “just” again

When most weight problems can just be fixed by just telling those people what's wrong with their diets, and actually offering them a simple solution, why not tell them that?

I'm also not American, and the people in my city I live in largely don't seem to have issues with not being obese, plenty also have food insecurity.

Food insecurity being equated to gaining weight is a hilarious leap in chain of events. Food education and upbringing, sure, but insecurity? Doesn't that mean food is harder to acquire?

elxr··on Novo Nordisk, Eli Lilly Report GLP-1 Signals of Slowed Biological Aging at ARDD
60 years ago obesity in every stable country was a tiny fraction of what it is today. Are people 10-50x less disciplined today? Really? Or are there just more incentives for processed food companies to sell you their crap? Are quick dopamine hits so easily accessible that people are becoming less driven? Is mainstream society not pressuring obese people enough to change course for being unhealthy?

Yes there's WAY more distractions today, and all those factors make it harder to navigate life practicing healthy habits, but you still have agency. The natural state of the human body is to use calories. The brain alone burns an astonishing amount of calories just thinking. Also no one's force feeding you anything.

There's plenty of countries with a lower "quality of life" index that barely have obesity problems. Maybe the natural tendency of the human body is not to slowly get fat?

elxr··on Novo Nordisk, Eli Lilly Report GLP-1 Signals of Slowed Biological Aging at ARDD
It's not hard.

If you can wake up on time everyday, shower everyday, brush your teeth, read emails; how hard is it to not eat obviously trash foods?

Nobody's asking you to learn a high cognitive load skill or play a competitive sport. You don't even need consider counting calories. Just stop eating things that aren't single (or few) ingredient fruit, meat, rice, potatoes, eggs, dairy, tomatoes, coffee, etc. Eat till you're satiated. Avoid added sugar / any sweet condiments.

No processed foods 6 days of the week. Jog for 10 minutes on days you're able to. Bodyweight squats. Short sprints if you enjoy them.

See how simple/doable those steps are? If you have the money for GLP1s and a fancy gym membership then by all means do whatever works, but acting like being fat in a first-world country is a difficult problem to solve is just strange.

elxr··on Android apps running native on macOS
At least it's short and every "no" clause adds critical information.

This one is fine.

elxr··on Novo Nordisk, Eli Lilly Report GLP-1 Signals of Slowed Biological Aging at ARDD
> Personal discipline of caloric deficit

Or you can just eat whole foods and do exercise. If you're an able bodied adult with a job, a basic understanding of the effects of foods with unnaturraly high sugar content, and even a slight care about your looks; then you really don't need personal discipline beyond what day-to-day life requires.

Being fat honestly takes more effort than being at normal weight for someone who's 100% able-bodied.

Yeah GLP1s work, but it always boggles my mind how eager people are to jump on them over the alternative.

elxr··on Show HN: 500 clocks in one 6.3 MB HTML file by Opus 5.5
80% of these are hideous. Why choose 500 instead of whittling it down a bit but having the quality bar a bit higher?
elxr··on Jev meets the Trolley Problem
Jev literally cannot write programs.
elxr··on Wayland desktop for Linux with skeuomorphic design
Love skeuomorphism. This just gave me a reminder to build a paper-like calendar app now, with pencil markings showing events.

I definitely don't want a skeuomorphic desktop, but little web-apps, text editors, chat apps? I say we need a lot more of this type of design there, especially with how good modern CSS is at this kind of thing.

elxr··on Apple HomeView
Hacker news isn't just programmers nowadays.

If this didn't have apps, but did have beautifully designed physical inputs, a color e-ink, and cost 1/3 as much I'd still be very excited for it.

It's just a bit derivative and isn't priced to take over the market (like the mac mini in 2025 was).

elxr··on Gemini 4 Argon
The model used is irrelevant when there's the critical issue where their "AI" assistant can't understand the word "nearest". Literally worse than a second-grade understanding of English.

I also have had shockingly bad results with gemini on android. Ended up giving up on it after a few months. Silly silly google with their trademark UX polish and attention to usability.

elxr··on Total Annihilation, rebuilt for modern computers
It's a sign of someone not caring about design.

One-shot websites from all the LLMs over the past 12 months (at the very least GPT, claude, glm, qwen based on my own experience) always add these stupid things. All you have to do is tell it to not include 2 digit section numbers in eyebrow text, and it'll stop instantly, but the people making these low effort sites just don't care.

elxr··on Show HN: BetterWispr – Free, open-source dictation for Mac
Sadly fluidvoice is mac only. People need to start pointing that out when recommending dev tools.
elxr··on "Just Ask Claude" Is a Rare and Valuable Skill
The fact that I can get (and consistently use) 120+ hours of 2160p/1440p video every month for $10, on a platform with extremely high quality, deeply researched videos/shows is a modern marvel.

Youtube music is also serviceable enough if you're not picky about mobile UX. Comes included with a basic yt premium subscription.

I'll always tell people, no other subscription - software or otherwise - comes close to providing the real-world value that yt premium does for that little money.

← PreviousPage 3 of 11Next →