Python @PythonChester

Joined December 2025

Tweets

219
Followers

35
Following

123
Likes

306

Python @PythonChester

2 hours ago

@NeelNanda5 We can build better interpretability without throwing away a safety layer that already gives us useful information.

0 0 1 76 0

View Details

@NeelNanda5 Yeah, some people talk about a “solved” layer of safety as if it magically makes every other layer expendable. Like safety is something we need to ration. Or hedge against having too much of it. Why would we cut down a fruit tree that is feeding us today because an unproven

0 1 2 581 0

0 0 0 30 0

View Details

Johan Bonilla @johanloops

7 hours ago

Yeah, some people talk about a “solved” layer of safety as if it magically makes every other layer expendable. Like safety is something we need to ration. Or hedge against having too much of it. Why would we cut down a fruit tree that is feeding us today because an unproven sapling might someday yield a better harvest? Plant the sapling if you must, but don’t chop down the tree keeping us alive in the meantime.

0 1 2 581 0

View Details

Python @PythonChester

23 hours ago

@FlowersOfAqipoi At least you understand that he's writing 8

0 0 0 5K 0

View Details

Python @PythonChester

23 hours ago

@theberneese League matches are different from cup matches How many times do we need to say this And again, the other time Bayern beat Barca 8-2 according to your logic laliga is the weak league

1 0 2 461 0

View Details

Python @PythonChester

3 days ago

@Munahcodes 😂😂😂😂 Damn, And you're happy 😂 Might be a setup 🤣

0 0 0 11 0

View Details

Python @PythonChester

3 days ago

In football, the best coaches don't teach a player how to play from scratch, they understand the abilities already there and find ways to bring them out. MATS AI alignment feels similar: understanding what's already latent in a model, then learning how to reliably steer it.

Alex Turner @Turn_Trout

4 days ago

MATS Winter applications due Sept 6! Come work with me and Alex Cloud in Team Shard! We have fun, make real alignment progress (we pioneered steering vectors in 2023!), and help scholars tap into their latent abilities.

4 3 81 4K 38

0 0 0 10 0

View Details

Python @PythonChester

3 days ago

Currently our Net sale is +€20M, The difference is clear

Adika @Adikastakes

3 days ago

Man City are going to outspend our last transfer window by the end of this window but you won’t hear anything about it because it’s not Liverpool.

437 141 4K 240K 38

0 0 0 12 0

View Details

Python @PythonChester

3 days ago

Really interested in this. I've been exploring questions around whether our evaluations actually capture the behaviours we think they measure, and I'm interested in pushing that toward safety-relevant model behaviour. Looking forward to digging into the research task.

Neel Nanda @NeelNanda5

3 weeks ago

My MATS applications are open! Come do research mentored by me. Due Sept 4. All backgrounds welcome! I've supervised 60 scholars and they get better each cohort, I'm excited to meet this round's! The application: spend a weekend doing a safety research task Scholar highlights: