miki123211

5 ms·

61.
by miki123211 2mo ago
Ah, so that's what Wolt (Doordash but in Europe) is doing. I noticed that Voice Over (iOS screen reader) crackles and randomly changes volume when using the app, but I attributed it to standard iOS weirdness, and possibly misuse of som
62.
by miki123211 2mo ago
It's funny how quickly we went from "the greedy US companies are subsidizing prices to keep competitors out of the market" to "the greedy US companies are overcharging because they are greedy."
63.
by miki123211 2mo ago
"are you sure" checks work if you force deliberate effort on the part of the user. Imagine having to type "I want to get hacked" on a keyboard layout which randomizes with every character.
64.
by miki123211 2mo ago
People also get scammed using accessibility services, so some banks deliberately make their app inaccessible unless you're running a whitelisted screen reader. If you are running a non-whitelisted screen reader...
65.
by miki123211 2mo ago
I swear, Semaglutide is reverse Asbestos. It's something we found and started using to one specific problem, but it later turned out that it has lots of other, extremely wide-ranging consequences. Except that in this case, these conseq
66.
by miki123211 2mo ago
Video isn't free to host, not at scale. Look at how much providers charge for "video on your own site" services. Sure, you can bypass these providers and implement everything yourself, but that isn't going to be much che
67.
by miki123211 2mo ago
Ads are progressive taxes in disguise. If you make the kind of content that costs a lot to create but almost nothing to copy, you have a dilemma: Set the price too high, and you aren't getting any money from the poor at all at all, eve
68.
by miki123211 2mo ago
> Nobody wanted us to spend six months looking for bugs in SQLite Most companies wouldn't. Instead, they'd fire the weirdo that came up with the idea of using some weird db, and switch to Postgres like God intended. I'm no
69.
by miki123211 2mo ago
Which is why you want magic links. Don't be the identity provider, have the email host be the identity provider (which it is anyway if you have a forgot password prompt). Agreed 100% that nobody should still be using passwords in 2026
70.
by miki123211 2mo ago
A lot of this is just weird Linux on the desktop weirdness. On Windows and Mac, VPN config is usually just installing a client, signing in, and then it all basically works. Sometimes you don't even need the client and can set it up in
71.
by miki123211 2mo ago
Is this because HS hasn't implemented it yet (which is on the community), or because Tailscale does something to prevent them from working specifically?
72.
by miki123211 2mo ago
I suspect that there are companies with internal proxies that load-balance across keys, and they didn't want to break that when adding encrypted reasoning.
73.
by miki123211 2mo ago
I'd lean yes; switching between models, even mid conversation, seems to be encouraged by harness vendors. No reasoning makes this much harder.
74.
by miki123211 2mo ago
IMO, !$ is most useful of them all (it is the one that gets replaced with the last argument of the last command). !! you can do just as easily with up-arrow enter, which is just as quick. !ssh you can do with ctrl+r, which actually shows yo
75.
by miki123211 2mo ago
Humanoids are the Turing machines of physics. They're Mediocre at everything, and worse than a dedicated machine at almost every task, just like a CPU is worse than an ASIC. However, they adapt to new tasks quickly. This decreases owne
76.
by miki123211 2mo ago
I really like the term "edutainment" for this. It's the kind of high-brow entertainment that makes you feel like you learn something. For me, most "push" things are edutainment, whether that'd be Youtube videos
77.
by miki123211 2mo ago
I don't think there's a ceiling to LLM capability. I do think that many software dev tasks are just far below that ceiling, and the gains to most dev work won't be that large from now on. Where the new generation of LLMs (Fab
78.
by miki123211 3mo ago
Keep in mind that this likely destroys context and makes your costs go up considerably. An agent often wants pipes so that it can `show_lots_of_logs | sed ...` and only see the part related to whatever error it's currently trying to de
79.
by miki123211 3mo ago
I have the opposite problem. I'm not well-calibrated on when I'd want lower reasoning than what's available to me (and how to compare that to lower-tier models). OpenAI now has Luna, Terra and Sol, each at Low, Medium, High
80.
by miki123211 3mo ago
Another great way to understand how vllm works is to read the code of nano-vllm[1]. It's basically "vllm but cut down to size. It's ~5kloc, supports just one model, disposes of some of the abstraction layers that vllm needs d
81.
by miki123211 3mo ago
In practice, I suspect the entire category will be folded into Claude Code and Codex (Anthropic is already trying with Claude Agents), and that will do 80% of the job at 0% of the cost and configuration effort. External programs will still
82.
by miki123211 3mo ago
On the other hand, we should recognize that, to put it plainly, not all dimensions on the frontier have the same level of importance. Any technological and societal progress is going to have negative consequences. To give just one example,
83.
by miki123211 3mo ago
The actual "prompting trick" that dramatically improves your results is often to add just two or three words, like "use library foo", "write in <language>", "<bar> algorithm". To know whic
84.
by miki123211 3mo ago
On one hand, I understand the frustration; and if I know I'm dealing with smart people who are pretty capable of Clauding themselves and who won't ask me if they can ask Claude, I would never do this. On the other... people really
85.
by miki123211 3mo ago
Oh, this is such a fun rabbit hole. There are plenty of programmers who can write a commit message in almost perfect English, yet don't know the difference between lettuce and radish, and can't tell what a little sheep is called.
86.
by miki123211 3mo ago
And to add to that, LLMs start with fresh context on each conversation, so they don't have the understanding of your system that human employees do. Nix lets you naturally encode that understanding in code (and in comments), making sur
87.
by miki123211 3mo ago
I think this often tends to fall along political lines. In my experience, the left tends to like simple stories ("we are poor and the capitalists are rich, so just take their money and give it to us", "if people love each oth
88.
by miki123211 3mo ago
I'm not sure whether that's the case. I think highly-educated people have a tendency to overestimate their expertise outside of their field, and especially in different subfields of the field they're expert in. Many (though c
89.
by miki123211 3mo ago
> OpenAI is demanding Just me, or are OpenAI and Anthropic basically hiring all the smart developers, like the Google of old? If you're making something like this, very likely that you're working for either OAI or Anthropic. Th
90.
by miki123211 3mo ago
Even if it is intelligible, reasoning styles (and hence reasoning effectiveness) differ between models. For example, gpt-oss loves reasoning in the style of "I be caveman, hungry, need food, need coconut, will search coconut now, eat w

More ›