Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
jwilber
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
24 ms
·
121.
▲
by
jwilber
3y ago
Way to throw a job away. How do they even interview for these positions?
122.
▲
by
jwilber
3y ago
Easily the best film I saw in 2023.
123.
▲
by
jwilber
3y ago
San Diego Pros: Best Mexican food in the country, proximity to Mexico, beaches, biotech (if that’s your industry), weather, laid back, great zoo, people are generally active. Cons: Very expensive CoL relative to income (“sunshine tax”: desp
124.
▲
by
jwilber
3y ago
Just letting you know: love the high-quality blogs you guys always write, especially the few on webgl!
125.
▲
by
jwilber
3y ago
If you enjoy interactive visuals of machine learning algorithms, do check out https://mlu-explain.github.io/
126.
▲
by
jwilber
3y ago
For a very high-level, visual introduction, see: https://mlu-explain.github.io/reinforcement-learning/
127.
▲
by
jwilber
3y ago
What? Even in that needlessly pedantic interpretation, their point still stands.
128.
▲
by
jwilber
3y ago
I think they’re speculating about the sort of person who is drawn to that sort of job, how they justify that line of work, etc.
129.
▲
by
jwilber
3y ago
5 years back I made a dataset of these, the USDA Pomological paintings, available on GitHub: https://github.com/jwilber/USDA_Pomological_Watercolors
130.
▲
by
jwilber
3y ago
For interactive articles on specific ai algorithms, checkout Amazon’s mlu-explain: https://mlu-explain.github.io/
131.
▲
by
jwilber
3y ago
“Whats utterly crap is the way large people are hounded into expensive seats on specious "it's your lifestyle" reason” I disagree. Airplane tickets are expensive and the seats are already extremely cramped. If you don’t adequ
132.
▲
by
jwilber
3y ago
“I don't think there's anything particular about SBF that makes him more likely to re-offend” No? Not the alleged (and very convincing) parent involvement? The total lack of remorse and continued denial? The latter is typical for
133.
▲
by
jwilber
3y ago
While many western tech salaries are high, the salary in discussion is a result of nepotism, not merit.
134.
▲
by
jwilber
3y ago
How exactly do you think taxes are used?
135.
▲
by
jwilber
3y ago
How this programming language maintains so much constant drama is both fascinating and concerning.
136.
▲
by
jwilber
3y ago
46 minutes later and it’s still on the front page. Does that change what you consider “telling” or not?
137.
▲
by
jwilber
3y ago
Here’s a visual, interactive explanation with d3: https://mlu-explain.github.io/neural-networks/
138.
▲
by
jwilber
3y ago
And here’s a shorter, visual explainer on cross-validation: https://mlu-explain.github.io/cross-validation/ What’s really interesting about k-fold cv is that most textbooks claim that the greater the value of k in k-fo
139.
▲
by
jwilber
3y ago
Would absolutely love this.
140.
▲
by
jwilber
3y ago
Awesome!
141.
▲
by
jwilber
3y ago
Notebooks are chiefly used for scientific exploration and experiments. The “literate programming” environment provides convenient artifacts for distilling research or analytics. Nowadays they can even be used for running models/analyt
142.
▲
by
jwilber
3y ago
Why stop there? I know plenty of PhD’s who would love to make bets with you about topics in their field - probably easy money for a smart guy like you!
143.
▲
by
jwilber
3y ago
What did you find ridiculous about their extrapolation? I didn’t find it ridiculous at all. Regarding your 8ft comment: plenty of footage has popped up of cops getting caught red-handed planting drugs. Much of it was filmed within 8 feet.
144.
▲
by
jwilber
3y ago
“… taken to a security room by American Airlines. There he was interrogated – and forced to buy a new ticket – according to the boy’s father. That’s all because when he checked in at the ticket counter, the agent saw his North Carolina driv
145.
▲
by
jwilber
3y ago
There’s an interactive neural network you can train here, which can give some intuition on wider vs larger networks: https://mlu-explain.github.io/neural-networks/ See also here: http://playground.tensorflow
146.
▲
by
jwilber
3y ago
“Once you've confirmed your email, we'll send the preview chapters to you right away.” Never received the chapters, but got signed up to a newsletter :/
147.
▲
by
jwilber
3y ago
This was the first open-source library I ever contributed to. Tom was super nice and way more helpful than he needed to be. Class act and great programmer/blogger: https://macwright.com/
148.
▲
by
jwilber
3y ago
“More than half the cohort claims to be part of the FIRE movement” Yeah, may have to double-check those poll results…
149.
▲
The Easiest Introduction to Neural Networks (2023)
(mlu-explain.github.io)
2 points
by
jwilber
3y ago
|
0 comments
150.
▲
by
jwilber
3y ago
Thanks, am definitely trying to squeeze them out as I’m able!
More ›