lysecret

8 ms·

91.
by lysecret 2y ago
The way I read it is they used ai as a coach and ai probably told them some variation of “it’s ok to exaggerate”. However, this to me would be a red flag because they somehow try to blame Ai for misrepresenting their experience. So they can
92.
by lysecret 2y ago
I used to do a lot of hiring interviews long before ai and this exact situation has happened many times. People have been added to some project doing x haven’t really done much or engaged in it. They then see you need someone doing x then t
93.
by lysecret 2y ago
I understand the scepticism here. For sure this app isn’t 90% accurate in any traditional sense. One note, as someone who also built a calorie tracking app with ai as well as lost a good amount of weight with it: accuracy for calorie tracki
94.
by lysecret 2y ago
I would appreciate a comparison to cloud tasks in your docs.
95.
by lysecret 2y ago
Ha this is a really good point! I worked with so many different kinds of observability approaches and always fell back to traced logs. This might be part of the reason.
96.
by lysecret 2y ago
Yea I also like this system only problem I was facing with it was http read will lead to timeouts/lost connections. And task queues specifically have a 30 min execution limit. But I really like how it separates the queueing logic from
97.
by lysecret 2y ago
This is awesome and I will take a closer look! One question: We ran into issue with using Postgres as a message queue with messages that need to be toasted/have large payloads (50mb+). Only fix we could find was using unlogged tables
98.
by lysecret 2y ago
My opinion. (Also as a German guy doing a similar thing). On the first order no. The tariffs just announced are only on goods and are collected at the ports when you physically import them. On the second and third order. What I think will h
99.
by lysecret 2y ago
Yep what many people overlook is that this deficient just means the rest of the world is financing the US overconsumption. And the only reason they are happy to do it is because they are happy to hold USD and are happy to invest in the US e
100.
by lysecret 2y ago
I worked a lot in F# and loved it. I love that it has a lot of great functional ideas without being too pedantic about being 100% functional all the time. (You can have mutating state or just call arbitrary C#.) I took a lot of its insights
101.
by lysecret 2y ago
Do you have a link to this talk by any chance?
102.
by lysecret 2y ago
I built this small website to retrieve personalized quotes/poems and questions also some habit tracking https://www.checkindaily.ai/ Also I have built a chrome plugin that can filter twitter by feeding your feed to gem
103.
by lysecret 2y ago
Really like uv too but surprised he doesn’t mention the lack of conda compliance. Some scientific packages only being available on conda is the only reason I can’t use uv (but micromamba) for some projects.
104.
by lysecret 2y ago
Google search example is interesting I wouldn’t care at all. YouTube or maps on the other hand…
105.
by lysecret 2y ago
If you know the exact tensor shape of your data ahead of time Zarr works well (we use it as the dataformat for our ml experiments). If you have dynamically growing data or irregular shapes zarr doesn't work as well.
106.
by lysecret 2y ago
I feel like there is already a movement, "thinking from first principles" along this direction.
107.
by lysecret 2y ago
I think the next big problem we will run into with these line of reasoning models is "over-thinking" you can already start to see it. Thinking harder is not the universal pareto improvement everyone seems to think it is. (I unders
108.
by lysecret 2y ago
Yea x and YouTube. Also the YouTube meta specifically seems to have moved to these sensationalist over exaggeration and over simplification. Really would be nice to have a sensibility filter somehow. It’s gonna come with ai.
109.
by lysecret 2y ago
I’m he other reason I really like cursors whenever there is a type/linter issue you can fix it with composer and it usually works well.
110.
by lysecret 2y ago
Hadn’t had much luck with o3. One thing that came to my mind with these test time compute models is that they have a tendency to “overthink” and “overcomplicate” things this is just a feeling for now but has anyone done some study on this?
111.
by lysecret 2y ago
Where do you have this "capacity" limit from? I can get as many H100s from GCP or wherever as I wish, the only thing that is capacity limited are 100k clusters ala ELON+X, but what DeepSeek (and the recent evidence of a limit in p
112.
by lysecret 2y ago
That’s actually a very good explanation thanks!
113.
by lysecret 2y ago
Did not expect this, but I had some immediate use cases for it (daily habit tracking, self-improvement type things). So I'm curious where this goes.
114.
by lysecret 2y ago
Funny, he starts of dismissing an AI IDE to end with building an AI IDE :D (Smells a little bit like not invented here syndrom) Otherwise fascinating article!
115.
by lysecret 2y ago
100% this is what I do in python too!
116.
by lysecret 2y ago
Actually prefer new geminis too. 2.0 experimental especially.
117.
by lysecret 2y ago
If you take vc money that’s the deal.
118.
by lysecret 2y ago
Isn't that the whole point of quantum theory, it's quantized? Or did I get that wrong?
119.
by lysecret 2y ago
I can only speak for the weather models (since it is in my domain). The answer is that the issues are much more engineering and scaling/infra issues than theoretical issues, and google is good at engineering (or attracts people that ar
120.
by lysecret 2y ago
Fascinating the fake plant appearing in the tech review video.

More ›