Jack 🇺🇸

AI @anduriltech, Christian, optimist

[Jack 🇺🇸 @jack_dlm]

Men only want one thing and it’s disgusting

[Jack 🇺🇸 @jack_dlm]

@parkerjbeard Sim is an elegant verifier

[Jack 🇺🇸 @jack_dlm]

@cwamidon @aphysicist Tech bros struggle to compose a realistic and compelling vision for what comes after the singularity

[Jack 🇺🇸 @jack_dlm]

@DeltaClimbs https://t.co/72xIhy2TyF

[Jack 🇺🇸 @jack_dlm]

@ring desperately needs to offer one-time precision air strikes from reaper drones for home defense (as a service) to their offerings

[Jack 🇺🇸 @jack_dlm]

@shouldntspeak https://t.co/zEDvVtwntJ

[Jack 🇺🇸 @jack_dlm]

@ring desperately needs to offer one-time precision air strikes from reaper drones for home defense (as a service) to their offerings

[Jack 🇺🇸 @jack_dlm]

Just another thing in life that being a Christian fixes

[Shaun Maguire @shaunmmaguire]

People have no idea how sophisticated nation state activities are

If you are a target of opportunity for China it’s not unheard of for them to dedicate O(1000) humans dedicated to compromising you

This is what happened w their NYT hack in 2013, which was designed to send a message

They will compromise your friend’s friend’s friend to compromise you

And work down that chain until your friend invites you on a fun trip on a private plane to Mexico/Thailand/wherever (example)

Where some night you have a very boozy dinner and then you show back up at the house and some girls are there

You’re drunk and make a mistake and then you’re owned until you’re disposable

Nation state kompromat for “target’s of choice” is a system designed to find any individual’s weakness (you’re bisexual; a gambler, an alcoholic, drug addict, have a debt, adulterous, etc) and then force them into a situation of maximum amplification of this weakness

99% of people can be compromised using nation state capabilities and tactics

https://t.co/EO5JEArsZa

[Jack 🇺🇸 @jack_dlm]

TARS ❤️

[Sam AlbertX @SadAlbert10]

fictional ai alignment tierlist

[Jack 🇺🇸 @jack_dlm]

Correct

[Jack 🇺🇸 @jack_dlm]

@ilaffey2 Christianity is simulation theory

[Jack 🇺🇸 @jack_dlm]

What happens if you use a modern GPU for graphics

[Jack 🇺🇸 @jack_dlm]

Or maybe it’s just succumbing to the singularity

[Jack 🇺🇸 @jack_dlm]

The only truly bitter lesson-pilled company would be self-assembling nanobots. After all, wouldn’t deciding on a product be encoding human judgement?

[Jack 🇺🇸 @jack_dlm]

Private equity is just post post consensus venture

[Jack 🇺🇸 @jack_dlm]

AI will learn autonomy best in very cheap embodiments. Evals will get expensive and hard to write, you will want robots to learn from physical experience, and you can’t get that much experience on an exquisite platform.

[Jack 🇺🇸 @jack_dlm]

@SamsaraDurvasu1 @TuvaAI Cool work! It is amazing how hard of a problem this continues to be for frontier models!

[Jack 🇺🇸 @jack_dlm]

chatjimmy is still by far the most compelling technical demo I have ever seen. @taalas_inc is going places

[Jason Wolfe @w01fe]

HOLD SWARM

[Jack 🇺🇸 @jack_dlm]

@cwamidon Truth nuke

[Jack 🇺🇸 @jack_dlm]

Getting your post-trained models into production and making them make sense in the long term is a very difficult problem given generalist model progress. Going to be a lot of disillusioned customers of post-training companies in months/years to come. Best to gather your data and build your evals first.

[Jack 🇺🇸 @jack_dlm]

models and harnesses are searchable, evals are all you need

[Jack 🇺🇸 @jack_dlm]

Gemini is back!

[Sundar Pichai @sundarpichai]

I did say we'd be releasing our models periodically...introducing Gemini 4 Carbon, our next frontier model.

Following Argon's release just two weeks ago, Carbon shows the pace at which we’re pushing the frontier, it ranks #1 on the recently updated v5.0 of @ArtificialAnlys Intelligence Index at 59. It also leads across key cybersecurity evaluations scoring 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching.

We’re seeing significant improvements in long-horizon tasks even compared to Argon, and we’ve strengthened frontier safeguards.

As we did with Argon, we’re going to roll it out gradually + responsibly. Some trusted cyber defenders will get access starting this week. We'll be expanding it to more partners through our Fairwind Program, ahead of wider availability soon, so hang tight!