HN user

aomix

410 karma
Posts2
Comments116
View on HN
Transcribe.cpp 4 days ago

What good timing to spot this. I've been reading more and more people talk about bringing TTS into their prompting toolkit and wanted to give that a try. The idea of rambling brain dump into a doc -> edit pass -> send to the robot loop sounds appealing.

In my experience this has an obvious answer (yes). LLM’s yes-and whatever patterns are present in the code base no matter how insane. If you don’t tend to your code garden it will be overcome with weeds in an instant.

At work we had largely moved off of some old hardware but kept partial support for it in the code because normally that would be zero cost. But the constant overhead of bullying, bribing, and threatening the LLMs to leave that code alone was way more work than just excising it and reimplementing it if needed.

James Duane, the lawyer from the famous video about never talking to the police, wrote a book called You Have the Right to Remain Innocent. It’s a great read if you want to feel very sad.

I credit randomly deciding to take Latin in high school as a huge change in the trajectory of my life. It didn’t even happen all at once. Just a little different perspective on how words are formed. Oh they’re often combination of simple root words.

Right after I graduated the one Latin teacher they had retired and that was it.

Talking the problem to death with the AI before implementation is a nice zone for me. I feel productive, get good results out of the AI, and still largely understand the code. That’s the part of the AI revolution that I feel has made me a better engineer because I argue about design and architecture all day with a robot.

It started looking a whole lot like OpenBSD’s random number system. Private entropy pool from good system entropy seeds a ChaCha20 stream with random reseeds for forward secrecy in case of compromise. I think Linux is even more paranoid in the early boot environment where even in the presence of a seed file it prefers to get system entropy mixed in before confidently saying it can do crypto activities.

I find it interesting that the same process that played out in the forums to feed transition also took place in video games with dedicated servers to matchmaking. I joined a random Counter Strike server in 2005 and ended up becoming close friends with regulars on the server and I'm still in touch with some of them this day.

I’m not even an athlete but I hike fairly often and walk 3-4 miles a day with my dog. If I get a pair of so-so boots they’ll last less than a year. I got recommended boots by a friend that lasted me 6 years until I finally had to replace them with an identical pair this Christmas. Nothing has been an issue in the two weeks I’ve had them but when companies get gutted by PE their quality goes down sharply. If I have to find a new boot company I’ll be very sad.

I never watched the widescreen version of The Wire they put out years ago but now I'm curious again. That show was a bone deep 4:3 product and the show plays with it constantly. Here is an interesting breakdown that made me really appreciate how clever they got while trying to be pretty subdued with the cinematography on The Wire https://vimeo.com/39768998

I’ve been agonizing over getting the Framework Desktop for weeks as a dev machine/local LLM box/home server. It checks a lot of boxes but the only reason to look at the Framework Desktop over something like a Minisforum MS-A2 is for the LLM and that seems super janky right now. So I guess I’ll wait a beat and see where we are later in the year.

I wonder if you plotted layoffs cycles you’d see a big gaps for tech companies for the last couple years since RTO orders will have the same effect.

Reading about the different RTO rebellions is interesting. I know multiple people who have ended up in a half day schedule that they really like. All meetings scheduled before noon-ish, then head home to focus on work. These are all kind of against company policy but there’s a shield of willingly ignorant managers between the executives and workers. As long as the managers can say “as far as any metrics I have access to tell me my people are in the office the required days per week” they’re ok with it.

I saw someone describe python as “stressful” for this reason and I couldn’t agree more. It’s difficult to have confidence in any change I make or review. I need to sit down and manually exercise codepaths because I don’t get the guarantees I crave from the language or tooling. While with the small amount of Rust code I’ve written lately I could yolo changes into production with no stress.

We had serious reliability issues with USB versions that were only solved by switching to the D457 model with GSML connectors. Those needed other hardware vendors to build in support for it to work but once that was all done we never had an issue again. Happy times going from alerts daily of the camera dropping out to alerts never.

I was working in a college job for a pharmaceutical distributor when the New England Compounding Center outbreak happened. I think that dropped a nuke on the whole industry. We stopped dealing with compounding pharmacies after that because that whole thing was more like a matter of when not if.

I didn’t appreciate this until covid and wfh. I’m an introvert and am in my happy place sitting in front of a computer or with a book. But I was losing my mind and had to be actively social for the first time in my life. I can see a decade of living like it’s Covid turning my relatively healthy, relatively young brain into soup.

Leaning into stereotypes, the older women in my family did just fine in retirement because they just started doing social activities full time. If anything they retired and got busier. The older men sometimes did ok but usually did worse.

Everything I've read about pledge and unveil really admire the approach and the results but it didn't seem to have a big impact outside of OpenBSD. It took ~20 years for OpenBSD's CSPRNG to be re-implemented everywhere else maybe we're operating on a similar timeline here.

I’m more bearish about LLMs but even in the extreme optimist case this is why I’m not that concerned. Every project I’m on is triaged as the one that needs the most help right now. A world when dozen projects don’t need to be left on the cutting room floor so one can live is a very exciting place.

DeepSeek-R1 1 year ago

They use other models to judge correct-ness and when possible just ask the model output something that can be directly verified. Like math equations that can be checked 1:1 against the correct answer.