Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
thom
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
22 ms
·
301.
▲
by
thom
2y ago
I continue to love these books all these years later, and my kids have started getting into them (cheating and all). Personal faves are Robot Commando (where you choose from a variety of differently configured robots at the start) and Crims
302.
▲
by
thom
2y ago
o3-mini-high also gets the right answer, in addition to mentioning Janensch.
303.
▲
by
thom
2y ago
To me… To you…
304.
▲
by
thom
2y ago
A bit like the three body problem, I guess. The problem seems simple because it’s just four walls, but think how complicated the constraints are once you have an arbitrary shaped sofa: you shave a tiny bit off one corner, and now you can tr
305.
▲
by
thom
2y ago
Yes and I’m not saying you’re wrong to dismiss it - all that work has been passed down as immutable, static laws that people regurgitate to say some things are good and some are bad. The fact is, most viz exists to support specific tasks or
306.
▲
by
thom
2y ago
It’s so easy to measure how fast and accurate users are at completing real tasks in your domain using various viz. I’d be disappointed if the conversation was purely between rigid principles and artistry.
307.
▲
by
thom
2y ago
I don't see those things as opposed - you want teams aligned, you want people to be able to plan and collaborate both the short and long term. A good process left shifts as much as possible so teams start out fairly well aligned and do
308.
▲
by
thom
2y ago
For me the internal part of this is far more important than any external pitch. Storytelling is a key skill for keeping an organisation aligned. Every company I've worked for that felt disjointed wasn't because of a lack of struct
309.
▲
by
thom
2y ago
Gotcha, totally fair points. It is an inherently conservative approach to standards, to take workarounds and pragmatic solutions that are already available within the community, and provide slightly cleaner (perhaps debatable) baked-in vers
310.
▲
by
thom
2y ago
People are going to build plugin and component systems whether we like it or not. People have massive investments in C++ whether we like it or not. These two concerns often intersect in game engines, apologies if that felt off topic.
311.
▲
by
thom
2y ago
C++ reflection is now good enough that hopefully we’ll start to see more game engines using it to work out components and properties instead of weird macros. jcelerier’s work on things like Avendish really does feel quite fresh and modern,
312.
▲
by
thom
2y ago
The author shows encapsulation of global state elsewhere. I’m not sure why they wouldn’t use RAII for the log level stuff so it was automatically rolled back.
313.
▲
by
thom
2y ago
Microsoft created a managed, wire-compatible alternative: https://github.com/microsoft/garnet
314.
▲
by
thom
2y ago
I have a different body but a similar relationship to it as the author. My wife is very stylish and often wears striking jewellery. She gets a lot of compliments and doesn't have to work hard to be at ease socially, whereas for the mos
315.
▲
by
thom
2y ago
Ah, the time I’ve spent in Tring.
316.
▲
by
thom
2y ago
I agree. Fascism is fascism whether it’s a teenage Octavian, Agrippa or a groyper. Underestimate them at your peril.
317.
▲
by
thom
2y ago
Very importantly here they provide a ways of decoding the encoded thought tokens, so you're not really losing explanatory power or debuggability. As much as OpenAI want to present hidden chain of thought as some sort of long term advan
318.
▲
by
thom
2y ago
The distilled versions are terrible and the full version seems very slow, but the latter is certainly in o1’s league.
319.
▲
by
thom
2y ago
Whether it’s data or things that can kill you, you’re guaranteed at least once delivery either way.
320.
▲
by
thom
2y ago
Wow, just when I’d accepted MIPRO in DSPy was magic, here we are. Things continue apace.
321.
▲
by
thom
2y ago
Terrible time to open a shovel store, amazing time to pick up a shovel.
322.
▲
by
thom
2y ago
Hey, some of them are working tirelessly to ruin sports.
323.
▲
by
thom
2y ago
I thought it was less that you could predict across packs and more that you could infer what card had been taken given what was left. That meant you had a better chance of not getting cut during the draft.
324.
▲
by
thom
2y ago
Used this for years and years and it’s always worked perfectly. Hugely appreciate that my license is still honoured after all this time.
325.
▲
Commoditize Your Complement
(joelonsoftware.com)
3 points
by
thom
2y ago
|
0 comments
326.
▲
by
thom
2y ago
This is the thing I specifically use LLMs for when I’m doing history courses. I’ll remember some vague quote or event and ask for the primary sources and latest ChatGPTs are excellent and getting the right reference, which I can then look u
327.
▲
by
thom
2y ago
If you could conjure up any data or video product from thin air to make your life easier, what would it be?
328.
▲
by
thom
2y ago
That's extremely good for a job in soccer analytics. It's a really fun domain, you'd be working with very smart and bought-in people at Arsenal, and they have their own collection operation so control their own destiny in a w
329.
▲
by
thom
2y ago
Fairly sure the vast majority of Hacker News isn't making 300k a year in either big tech _or_ startups. Plus the author of the post mentioned wasn't some scrappy gung-ho founder trying to make it at any cost, they had Google money
330.
▲
by
thom
2y ago
Hashtags became links around 2009, but I think it was just critical mass. Instead of yelling into the void, it became very easy to stumble upon a community or discussion around a hobby or event, and follows didn’t require approval like frie
More ›