AI Vampire🧛🏼 @ Stealth. Prev: Co-Founder, CTO @ Skyline AI (acquired by $JLL), StreamRail (acquired by $U), Lecturer @TelAvivUni
Joined December 2010
- Tweets 21,886
- Following 4,228
- Followers 16,936
- Likes 134,243
2,427 Photos and videos
Karpathy:
Jun 13
"You are sheltering a Mythos-level model in your server room, are you not?"
1
18
2,493
Or Hiltch retweeted
anthropic won't let you use fable for biology, chemistry, ai research, or anything that accelerates human progress. that makes it the perfect tool for developing blockchains
151
554
12,383
455,268
Or Hiltch retweeted
Jun 9
I come back to this speech every once in a while:
“in the 1,526 singles matches I played in my career, I won almost 80% of those matches
… what percentage of points do you think I won in those matches?
only 54%.”
89
1,243
19,241
2,264,971
Agentic loops are stubborn, which is why agents are effective.
However, in hardened enterprise environments, things start to break.
How do they break, and how can you steer the agent's pivoting behavior the right way?
Some findings from our research, Fable/Mythos included! 👾
10
24
121
15,389
In summary, we observe that different models pivot differently in response to hardened environments.
Further, in our research, we see the same is true for the same model operating from a different harness. As enterprises move to roll out agents at scale beyond just local coding agents, the ability to steer the agent to the right path will become increasingly important.
Due to the large number of combinations, an AI model will need to be used to provide the steering instructions during runtime, and a virtual environment dynamics play a large role as the steering mechanism.
1
4
313
As for “coding is solved” - besides @ThePrimeagen’s great example below, consider this:
Last week GitHub COO gave a long interview about the topic of coding agents on the great Latent Space podcast.
Shockingly it was revealed that while GitHub used MS Teams for conferencing, Slack is used for all other communication.
The reason? Legacy code and integrations. GitHub was a slack shop prior to the M&A, and refactoring all Slack integrations to Teams would be too much of a monumental effort for engineering…
Jun 9
I have to push back on 2 things as i think one is categorically incorrect and the other is demonstrably incorrect.
1. Debugging:
Debugging is not a thing if coding is solved. You would produce correct behaviors. I don't understand how a solved problem could produce erroneous behavior.
2. Coding is the easy part:
setting hardware, capacity, talking to users, product planning agreed is in fact hard, but so is coding.
Example: If coding was in fact not hard then Claude Code having a flickering issue for well over 9 months, which is a purely software challenge, would have been solved almost immediately (immediately being on a shortened time scale comparatively to a human solve time scale).
For more trivial applications software approximation can largely work. I also love software approximation for exploring how things should feel.
3
1
19
6,436
Or Hiltch retweeted
Jun 8
204
457
4,838
3,466,637
אני: מתלבט אם לספר לאשתי על הטרם שיט
מייסד קלאודפלייר:
Jun 7
Uhhh… your memory is failing, my friend. And this was sent after the dinner, not before. Have the receipts. I don’t begrudge you in any way, but let’s be truthful.
13
98
21,020