codegraph ๐Ÿ‡บ๐Ÿ‡ธ (@codegraph) header
codegraph ๐Ÿ‡บ๐Ÿ‡ธ (@codegraph) profile picture

knowledge engineer

๐Ÿ‡บ๐Ÿ‡ธ
Joined April 2021
Replying to @craigweiss
The hiring manager said I had too many PRs.
1
1
78
The thing in the back of my mind nowadays is SFRE. It's national guard special forces tryout for civilians and military. It would almost certainly be bad for my life to do but I'm gauging my runs and fitness on if I could pass or not.
1
1
48
Replying to @graphtheory
By the identity property, the floor is on the floor.
1
3
34
codegraph ๐Ÿ‡บ๐Ÿ‡ธ retweeted
Replying to @warDaniel47
POV: youโ€™re listening to a supermodel staff meeting after a 12 hour outage call @codegraph
1
1
53
codegraph ๐Ÿ‡บ๐Ÿ‡ธ retweeted
sick video from mastercam on the carmelite monks in wyoming using 5ax machines to build a gothic monastery imagine rolling up to the shop and your programmer is brother peter thomas youtu.be/wLJa410KTsg?is=pun-โ€ฆ
41
117
25
1,638
36,393
Replying to @forthrighter
You're writing for AI more than people now. You need to be consumable for an LLM to get rolled into the training data and pushed to a wider audience. Serial blog makes the most sense in this case for context window purposes.
2
28
Replying to @EdCarney172811
Pope only has 2 rounds
1
29
2,151
Replying to @helicitywatches
This is my favorite comment because I've actually been buying new shirts for these interviews.
3
239
4,305
Replying to @say_HelloMax
Roughly 1 per week with feedback within 3 business days and another business day to schedule the next one. One round had a 2 week gap and multiple interviews in one day.
9
39
5,462
Replying to @Flower_of_Babel
Sorry that comes after the 9th round
2
214
9,373
Just got moved to the 6th round of interviews at a company. What are we learning at this point?
472
57
33
2,826
140,610
codegraph ๐Ÿ‡บ๐Ÿ‡ธ retweeted
now that we're interpreting monosemantic neurons it's time to bring this up again. fast memory recall & mechinterp are the same problem. let's figure out activation-triggered indexing. i think it's doable
imo e2e mechanistic interpretability has the highest chance of leading to e2e memory recall by providing activation-triggered indexing and search
1
3
70
2,874
Replying to @graphtheory
Noho is not THAT bad
1
1
29
codegraph ๐Ÿ‡บ๐Ÿ‡ธ retweeted
these qwen neurons are insane like wym i poke its brain and now it's an alien named Extrali who's been guiding civilization for 3000 years lmfaoo
72
146
45
2,749
63,325
codegraph ๐Ÿ‡บ๐Ÿ‡ธ retweeted
The perversity of hot dog vending must be refreshed from time to time by theology and geometry.
2
3
1
5
570