I feel like the “British food is terrible” meme was born from specific British-French banter and is now parroted by people from countries with food that is objectively as bad or worse.
mctoady.eth
Verified account Quantstamp@TrainTestToad
audit engineer // privacy enjoyoor // shitposter // toad
views my own ☕️🐸
views my own ☕️🐸
689 Following 1.3K Followers 5.6K Posts
Latest posts 5
The main thing I'm taking away from Devcon is that we need more spaces (on and offline) where smart people can get together and discuss interesting ideas.
If you could bottle and sell the energy in the air when passionate people are trying to solve hard problems, you'd be rich.
If you could bottle and sell the energy in the air when passionate people are trying to solve hard problems, you'd be rich.
Working out how we can provide a path for junior SRs in the age of LLMs was a recurring undercurrent of conversations I had at DSS/Devcon.
It is easy to very quickly appear to be relatively productive in audits by using LLMs. But if you are essentially a meatproxy communicating the output of whatever frontier model you are using, your ceiling is capped and you aren't going to find anything that a competent client wouldn't have found during their own AI review.
On the other hand, if you do things the "right" way you will likely be worse (and 1000x slower) than an LLM. Then, the question becomes how do you keep yourself afloat for X months until you cross the LLM baseline?
This is an open problem in the space and unfortunately I have no silver bullet to offer in this post. I'm going to be reaching out to some educators in the space for ideas, but I would also be interested in hearing from upcoming SRs about how they are approaching this and the difficulties they are running into. 🩵
It is easy to very quickly appear to be relatively productive in audits by using LLMs. But if you are essentially a meatproxy communicating the output of whatever frontier model you are using, your ceiling is capped and you aren't going to find anything that a competent client wouldn't have found during their own AI review.
On the other hand, if you do things the "right" way you will likely be worse (and 1000x slower) than an LLM. Then, the question becomes how do you keep yourself afloat for X months until you cross the LLM baseline?
This is an open problem in the space and unfortunately I have no silver bullet to offer in this post. I'm going to be reaching out to some educators in the space for ideas, but I would also be interested in hearing from upcoming SRs about how they are approaching this and the difficulties they are running into. 🩵
Now I don’t have a marketing degree, but has giving away free tshirts at conferences ever produced a long term positive outcome for a project?