quickthoughts

5 ms·

1.
by quickthoughts 3mo ago
Ha just wrote a post[1] about a sort of 4th option - max out cheap compute to create more tangible things that can be used/run locally. 1: https://news.ycombinator.com/item?id=48519181
2.
by quickthoughts 3mo ago
what does this actually do? claude output -> drop -> app that you can't really change?
3.
A case for AI-pragmatism – leverage cheap compute while it's here(quickthoughts.ca)
4 points by quickthoughts 3mo ago|0 comments
4.
by quickthoughts 3mo ago
Are most corporations preformative? (yes)
5.
by quickthoughts 4mo ago
I use small models like Gemma to improve transcriptions from ASR models amongst other micro-tasks. I actually built out a fine-tuning whisper pipeline with all local (smaller) models meaning no cloud/big-tech co is able to train/s
6.
by quickthoughts 4mo ago
I don't think this will be the case, image and video processing typically is more computation hungry, especially in realtime contexts. Apple already pegged Google as their cloud-provider of choice when it comes to LLMs/AI compute.
7.
Apple's privacy invading tech is likely going mainstream(quickthoughts.ca)
6 points by quickthoughts 4mo ago|6 comments
8.
by quickthoughts 6mo ago
It's everyone. Especially google, but all the big tech companies play in the same pool. Amazon, Google, Apple, Meta etc make money selling ads, which ultimate enables the tools that result data harvesting from everyone across the inter
9.
by quickthoughts 6mo ago
I've been working on the flip side of this with ASR models, but the problem space is the same, conversational/real-world data is needed. Whisper often mistook actual words I say and hallucinate all the time when speaking technical
10.
by quickthoughts 1y ago
Plus one for the IKEA pour over. Its cheap and functional, perfect for travel too.
11.
by quickthoughts 1y ago
VSCode with auto complete and Gemini 2.5 Pro in a standalone chat (pick any interface that works for you, eg librechat, vertex etc). The agents-in-a-IDE experience is hella slow in my opinion. Plus its less about the actual code generation