roncesvalles

It's unusable only because I can't set 5.6 to Extra thinking in the "Chat" tab, and the "Work" feature that does let me use Extra is totally gimped for coding tasks. I'd rather use Opus 4.8 Extra/Max instead of Sol High (look up the benchmarks, thinking level is everything).

It is quite literally unusable to me.

>5.6 Sol, competitive with Fable (or close enough), and it's available via subscription (even the $20 subscription!)

It's not comparable because OpenAI caps thinking to High in the ChatGPT "Chat" interface (and the "Work" thing where it actually does let us use Extra or Max is fucking shit).

GPT 5.6 Sol (High) is almost certainly worse than Opus 4.8 (Extra), and nowhere close to Fable (Extra).

I literally got a refund for my $20 OpenAI subscription after playing around with 5.6 Sol for a couple of hours (yes even on Codex) because it's so unusable and I'd rather just use Fable today and 4.8 Extra starting tomorrow, still within my $20 Anthropic plan. And I'm not even poor.

I still think the best way to build software using LLMs is to copy-paste snippets/files into the chat and manually guide the work. Humans are still the best orchestrators. Yes the human has to now be hyper-focused and juggle various workflows, but the end result (quality of work and throughput of usable code) becomes very good.

Wails is another cross-platform GUI framework in Go: https://wails.io/

Not to forget the whole thing with Free Basics. Zuck is smart enough to know why trying to subvert net neutrality in developing countries is a bad thing, and he did it anyway.

Of all the vilified tech billionaires, Zuck is the only one that I actually suspect of being "evil" -- he knows exactly how his business decisions are bad for society, he just doesn't care. He does business pure minimax style, unwavering focus to fiduciary duty, and that's likely why he was able to build Facebook/Meta into what it is.

I'm guessing if you can "see" Wifi through a wall, you can also infer some things about what's behind the wall, because the signal will have certain obstruction patterns. It becomes an xray.

It seems in mathematics that the utility of a problem is directly correlated with how difficult it is to solve, for some odd reason. If I defined some pointless construction and it turned out to be very difficult to prove, it would automatically over time become considered a "high utility" mathematics problem (again, for some odd reason).

Mathematics is largely just smart people working on pointless puzzles, and only by coincidence do these puzzles turn out to have practical applications (it cannot be predicted). Or I guess all the obviously practical problems in mathematics have already been solved -- we're now in a world where math is rarely the limiting factor for human progress (like it was, say, pre-calculus; was FFT the last significant unblock from math?).

It's such a waste of the best human minds. Or maybe the best human minds are actually doing something else, maybe we only notice the handful of Terence Taos, not the hundreds of people of equal brilliance who realized pure math is pointless and decided to pursue physics, rocketry, or quantitative finance.

I think it is widely known by now that instructions to alter the LLM's "tone", things like asking it to adopt a persona ("you are the world's best programmer"), and overly broad directives ("make no mistakes") always gives poor results. Just state directly what you want. If you want something very specific, add more information. "Prompt engineering" is pseudoscience.

To put it another way, you will only get the benchmarked performance if you let it talk the way it talks by default. Trying to modify this neuters the model's IQ.

Also consider that Fable launched on June 9. Many people including myself bought a 1-month Anthropic subscription just to use Fable.

By extending it to July 12, they're gonna get a second month out of a lot of such people. If it really expired today, I wasn't going to renew my month.

On this note, (from a web dev perspective) I think we've overdone making the static build as lean as possible and pushing everything into the backend. E.g., if a 2MB dataset will make a search/select bar return results instantly instead of having to roundtrip an API, just send up the whole dataset with your initial load. After gzip it's nowhere close to 2MB anyway.

Also, I think suspense is an anti-pattern now. We've all been conditioned to groan every time we see suspense animations because it means the website must be slow. It's better to just have a little circling loader or just freeze.

There is a pop theory that aging is just scurvy in slow motion and Vitamin C hypersupplementation actually goes a long way in staving it off.

To be pedantic, master-master implies the same row can be committed to at the same time at different master nodes (without an immediate causal connection between the two). Very few databases allow this.

You need replication for HA. Otherwise the only HA you have is the RAID setup on the one machine. The reason people go horizontal is rarely scale. And if you're doing that you might as well use the passive replica as a read replica.

It is a search engine in the sense that it's returning to you a subset of knowledge from a large body of knowledge stored within it. Natural language is simply the interface by which you make the query and by which it can return its results.

And true, it's not deterministic in our experience, but this is an optimization. You could make it so that one particular prompt always returns the same response every single time; the random jitter is "added in" because it happens to produce better results overall.

>They speak fluently and confidently about experiences it’s impossible for them to have.

But they're echoing these things from people who really have.

The key is to not forget that LLMs are just next-generation search engines, instead of anthropomorphizing them to be "speaking agents". The natural language IO interface is just a side effect.

The causality might not be what you think. You're happy with what you have because you don't see a path to getting substantially more.

I would say Europe in general. And I get the sense they don't aspire to social mobility either, the way Americans do.

there are only 2 and they don't seem to be AI-written

At many companies (I want to say most), Redis is seen as an actual durable production database and operated that way, not just as a cache that can disappear at any time. It's not unreasonable for a new dev to assume this unless told otherwise.

Yes but not very different results (unless you're adding new information to your prompt or reducing some ambiguity). Prompt engineering is mostly pseudoscience.

A lot of people who write code aren't professional SWEs. A lot of professional SWEs especially on Blind aren't typical SWEs at typical companies. A lot of people who claim to write code are lying. I mean would you take Gary Tan as a valid datapoint for how effective AI is at professional SWE coding? I wouldn't. But these are the kind of people who shill it the most on HN.

Also, I suspect that a lot of HN comments that generally state that AI has been useful for their work are actually not referring to programming work, but it's very easy to default interpret it that way. For example, I just recently saw a comment stating how AI can do 90% of their senior-level job, and when I looked into the guy's profile, he was actually a designer, which makes sense. But on first read I assumed the comment was referring to writing code.

You aren't allowed to block PRs for being too large anymore. The objective that every engineer should be 2x/3x/5x more productive can only be achieved if you go totally lax on code reviews.

Because if all your SWEs produce 5x more code, it also means they have to review 5x more code. But LLMs don't really help with code reviews. Then it becomes a Metcalfian paradox unless you just rubberstamp PRs, which is what is expected of you.

>This is mostly due to incredible pressure from the C-level for every engineer to be using as much AI as possible

I think this is an important point. Software engineers always had the right instincts on how to approach AI for coding -- cautiously. Execs got too coked up on LinkedIn puff pieces from nobodies and adver-prophesizing CEOs selling their tokens and chips that they forced something unnatural upon their orgs.

Now what we see in the software dev space is incredible levels of malicious compliance ("you want slop, I'll give you slop").

It was never the same quality though. A VM at AWS or GCP is not even remotely comparable to a VPS at some low-end provider. Things like RAID, network/powersupply redundancy, internal privacy controls (nothing stopping a VPS provider from snooping in your filesystem), software patch quality, noisy neighbor mitigation, bandwidth quality, SLAs etc. Then all the value-added stuff like Terraform support, logging, monitoring, IAM roles, instant scalability etc. Not to mention the benefits if you use the same cloud's managed services like databases.

If I can get a same-spec EC2 instance at AWS for just 2x the price of a Hetzner box, I would never consider Hetzner. It had to be a "10x savings".

OS X was never good and still isn't. I need to install like 7 third-party usability tools to make myself productive on a Mac, including a tiling manager, a clipboard manager, and a not-broken replacement for Spotlight - very basic things.

Also, Steve Job's "font obsession" is overrated. The fonts on Windows have always been much better and render way better as well (even to this day). Helvetica Neue is widely considered one of the worst fonts and Apple used it for a whole 3 decades.

This is looking back at Apple through some very rose-tinted lenses. Apple had a big role in moving the tech industry toward "grift-adjacent". There were at least 2 contiguous decades when Apple products were unusable, poorly designed, self-important, overpriced pieces of junk. Some would argue they still are.

People bought Apple because they were subscribed to Steve Job's personality cult. Heck, they might've even bought a "not-a-flamethrower" if he tried to sell one.

Not really. Firstly, speed and access are actually money problems.

Secondly, motivated state actors have no dearth of money. It's not like the Chinese are saying "oh it would cost a bit too much to infiltrate all US financial systems, we'd have done it if it cost less."

If it can be done, it would've already been done. No party that matters is materially "unblocked" by the advent of LLMs.

There is nothing that LLMs can do that humans cannot. If you are worried about bank accounts getting hacked, that's a problem to be solved on the banking side.

>Troubles for Anthropic IPO would channel all those money into OpenAI's one.

Troubles for Anthropic would almost certainly affect OpenAI, significantly. Yesterday just proved that the government sees it within their remit to shut down AI models. All current and future AI investment now has to contend with this risk. You should even see the effect of this decision on SPCX on market open despite X.ai being whatever tiny fraction that it is.

There is nothing to indicate that LLMs are improving "exponentially" at this point.