Ultimatt

10 ms·

61.
by Ultimatt 4y ago
So 7 cents for dozens of requests is only about 1/10th what I was saying. So could be I have the old API, but even that 7 cents for 10s of requests is not cheap compared to executing a model yourself at scale.
62.
by Ultimatt 4y ago
Sure but at some point you're testing prompt generation and what happens with the model, thats what Im talking about. This is a basic session with a couple of people clicking cost that much. So clearly Im doing something far more wrong
63.
by Ultimatt 4y ago
Youre reading the snarky gusto in your head. My point was literally that small mistakes and even something operational scaled beyond extremely small user bases is not "cheap". If two humans clicking is five bucks in an afternoon.
64.
by Ultimatt 4y ago
Could be I used an older API with the newer model. But there was no loop around the request only human input with mouse clicks from two people. Whatever was happening on the billing side there is zero chance Id ever post a project to HN for
65.
by Ultimatt 4y ago
You're right it was $3.74 https://ibb.co/gRLQBZ4 Returning under 300 characters and the prompt sent was about 10 words in length.
66.
by Ultimatt 4y ago
I suspect the author doesnt realise one request with hardly anything returned is many hundreds if not thousands of "tokens". It adds up very fast. Just some debug effort on a nonsense demo learning project cost $5 in a couple of h
67.
by Ultimatt 4y ago
Oh man Phoenix those were the days, 2003. I imagine a lot of HN readers here were only just born.
68.
by Ultimatt 4y ago
This problem feels like a no runtime type problem. Highly dynamic languages have tonnes of grim problems like this, that they have to deal with because there is no good type information anywhere so stats at runtime is how you optimise. Raku
69.
by Ultimatt 4y ago
You are underestimating how easy it is to email a lab service to synthesise a plasmid for you... Just first hit from google is this place in the EU https://eurofinsgenomics.eu/en/gene-synthesis-molecular-biol... But yo
70.
by Ultimatt 4y ago
The question more generally how easy is it to engineer bad stuff, regardless of this kit. Easier than engineering good stuff!
71.
by Ultimatt 4y ago
That would be a very bad idea, as what you break is fundamental reality around you. A lot of people on this thread really underestimate the level of care that's currently taken with the sorts of labs that are used and how controlled th
72.
by Ultimatt 4y ago
I should add those 3 months would also pay me about 2x better than a postdoc working fulltime in a university. Thats the level of pay disparity with industry, and the level of pointlessness of a university as an institution if you dont need
73.
by Ultimatt 4y ago
The point you missed is she got where she is with a very very large amount of luck too. Most grants have about 100+ equally high quality applicants at this point in history. If you dont make it 8 years out of PhD (which is a huge investment
74.
by Ultimatt 4y ago
Plenty of research doesn't need millions to do. If you're a mathematician or almost any kind of theoretician or computational scientist you need a laptop maybe the odd bit of cloud time on a bigger blob of compute, in the case of
75.
by Ultimatt 4y ago
"WiFi Routers Estimates 3D Pose of Humans in Modelled Reconstruction" as an actual non bullshit title. The paper being a more concise "DensePose from WiFi".
76.
by Ultimatt 4y ago
This is definitely good advice for fleshing out the actual writing, I used this ontop of to the paragraph planning. So you build out a mindmap style plan with part > chapter > section > sub-section > sub-sub-section / figur
77.
by Ultimatt 4y ago
also need STALE and STILL-WARM
78.
by Ultimatt 4y ago
Go read what I said, the premise was you can afford the item outright but it still doesn't make sense to if you have interest free credit available.
79.
by Ultimatt 4y ago
Dude what world (or country) do you live in? Im in the UK and everyone has a £1k self employment allowance, that gets you quite a nice camera! Filling out the return is trivial.
80.
by Ultimatt 4y ago
Because I assumed you'd understand the depreciation of the asset is a lot less than the appreciation on your investments or cash that you can do in the other scenarios. The entire point is its offset. Also that's not a business as
81.
by Ultimatt 4y ago
You then pay off your outstanding debts if you think there is risk you will be paying out more in the near term... That's why you only carry debt when you could otherwise afford something. Literally thats the point. Realistically thoug
82.
by Ultimatt 4y ago
Or you can afford the camera outright, right now. What's better out of these options: 1) Buy for cash the camera now. You don't have any cash in hand afterwards, and the camera is depreciating in value immediately as an asset only
83.
by Ultimatt 4y ago
Klarna does more than just BNPL loans though, they do normal credit too.
84.
by Ultimatt 4y ago
The implementation is optimised for genomics, thats why.
85.
by Ultimatt 4y ago
50ms doesn't sound especially fast. If you had a million of them to process thats 4 hours, maybe an hour with some parallelism? A 100x speedup from going native then becomes quite welcome. Python has its place, but the fact you even th
86.
by Ultimatt 4y ago
This blog article was missing the first step of using Numba on numpy https://numba.pydata.org/
87.
by Ultimatt 4y ago
I prefer not_sent = users - {u for u in users if send_welcome_email(u)}
88.
by Ultimatt 4y ago
The space of startups cashing in on genomics but making shiny web apps that software engineers need to understand something about human diploid genetic variation is far higher though. Thats where the money and engineers are, not in fundamen
89.
by Ultimatt 4y ago
The opening paragraph of this resource states its absolutely not about being a comprehensive introduction to genomics. I strongly disagree with the sentiment its naive or oversimplified. It's trying to give someone with no knowledge a
90.
by Ultimatt 4y ago
Share holder dividends. RELX is 2.21% Not sure why everyone forgot capitalism in this.

More ›