Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
scollet
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
13 ms
·
151.
▲
by
scollet
4y ago
Hm. You know the answer. If you don't, walk out your front door.
152.
▲
by
scollet
4y ago
Ah I did not know.
153.
▲
by
scollet
4y ago
You didn't have to post a novel. It ends here: "Free means it's rightfully theirs".
154.
▲
by
scollet
4y ago
They are not. Companies are abusing data. The services should still be free.
155.
▲
by
scollet
4y ago
Can we get a venn diagram of people who write for thehustle and people who regularly snort cocaine?
156.
▲
by
scollet
4y ago
Don't most production NN or DLN optimize to a maximum? Seems like the behavior becomes predictable and then you have to retrain if you see unoptimal results.
157.
▲
by
scollet
4y ago
Even with those limitations, doesn't metal programming optimization compound in later abstractions? Seems worth the investment from a rational standpoint.
158.
▲
by
scollet
4y ago
Seems like a linked list (and set dressing) is circumstantial. Is it because linked lists can be slow and can sometimes pocket arbitrary data? I feel like strong typing and limiting pointers in favor of runners combats this.
159.
▲
by
scollet
4y ago
I should have included "also".
160.
▲
by
scollet
4y ago
When job requirements actually match the job, then you can worry about this.
161.
▲
by
scollet
4y ago
Twitter is about to become a marketing platform.
162.
▲
by
scollet
4y ago
Unicorns are DoA in my opinion. If everyone's a unicorn, no one is.
163.
▲
by
scollet
4y ago
Shouldn't you do this in a bull?
164.
▲
by
scollet
4y ago
I think we will see even more optimization with constrained budgets. Software is pliable. You will have a network of middlemen running at cost.
165.
▲
by
scollet
4y ago
You would have better luck running the comments through GPT.
166.
▲
by
scollet
4y ago
There is still an element of optimization which involves compressing the problem space. This might be why we see thin vertical segments, but they are way over-valued for the optimization they provide. I think VCs' expectations misalign
167.
▲
by
scollet
4y ago
The dark pattern for me is seeing ads.
168.
▲
by
scollet
4y ago
The final panel is the removal of the fence. You're conveniently leaving that out.
169.
▲
by
scollet
4y ago
Easily top two worst nightmares is being in the presence of a large cat or bear and the thing is looking at me. It's horrifying.
170.
▲
by
scollet
4y ago
You are the system understanding itself. Your brain is distributed.
171.
▲
by
scollet
4y ago
It's possible that there never was a first place.
172.
▲
by
scollet
4y ago
There is some meta humor here about the classifier not understanding absurdism.
173.
▲
by
scollet
4y ago
Just straight up linking to a dog whistle sub... okay...
174.
▲
by
scollet
4y ago
Adderall does not work on NT brains. They would be able to perform tasks regardless.
175.
▲
by
scollet
4y ago
You don't even need to know how to grow to a potato, you just need to be able to learn how to grow one.
176.
▲
by
scollet
4y ago
Dad said I could stay up late!
177.
▲
by
scollet
4y ago
Well that's the problem, is that the incentives for instruction are misaligned and the ramp to excellence are ill-defined. The boot camps are wrong in teaching rotary, where they might benefit from selective regiments. Learn to learn a
178.
▲
by
scollet
5y ago
I personally do find those problems fun, or some of them. They rarely crop up in a professional capacity. I've mostly seen them in my hobby gamedev work where things like concurrency and graphs and dynamism are endemic. That's why
179.
▲
by
scollet
5y ago
Bezos does not create wealth though.
180.
▲
by
scollet
5y ago
> Why isn't he in jail, I mean, that's the hope.
More ›