fofr

Whispering to models at Google DeepMind

[fofr @fofrAI]

Argon has this amazing natural understanding of the world. It can do so many things from first principles.

Here's a pelican riding a bike, using fourier epicycles.

Video made by Argon.
Time-synced Lyria 3.5 music prompted by Argon.

[fofr @fofrAI]

Gemini 4 Argon is live.
https://t.co/SP1fiLpCWW

There's a lot to show you.

[fofr @fofrAI]

Gemini 4 Carbon shows a 9 point increase over Argon, which we released 2 weeks ago.

[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!

[fofr @fofrAI]

Announcing Gemini 4 Carbon 👀

"Carbon illustrates the pace at which we’re pushing the frontier forward. Our researchers and models are working hand in hand to advance capabilities faster, while strengthening our safeguards in tandem."

https://t.co/MtqIxZytCH

[fofr @fofrAI]

@simongmrk Are there some known good examples of long chaotic phase stars?

[fofr @fofrAI]

Some more context on the weird star, Gaia DR3 4582145718700356096 ("fofr's star"). It's an old star, and according to our planet formation models, it really shouldn't have very large amounts of dust.

If it's dust, "it holds ~100,000x more dust than a system that old should"

[Adam @lowgmath]

This gets weirder.

I searched ~750,000 Sun-like stars for the signature of Gaia DR3 4582145718700356096. At least two more have it, glowing right on the star:

J000153 (Gaia DR3 2874265305022213760)
J144944 (Gaia DR3 6225457312033584384)

All three look normal in visible light but shine 50-100x too bright at 22 microns, at 200-280 K, about Earth to Mars distance from each star.

If that’s dust, it’s at least a 200–300 km asteroid ground to powder. Dust like that usually comes from young planets colliding. This is what happened with our Moon.

But Fofr’s star isn’t young. It’s probably about as old as the Sun, old enough for a planet to have carried life for billions of years. And it holds ~100,000x more dust than a system that old should, based on the planetary formation models we have.

The best-known old case, BD+20 307, has a twin star circling it every 3.4 days, which may stir its planets up. None of these 3 stars have any companion.

So either our picture of how planetary systems age has a gap, or this isn’t dust at all, and something else is turning 1–3% of these stars’ light into heat.

[fofr @fofrAI]

@chrisfirst I've seen this black mirror episode.

[fofr @fofrAI]

@nikodemszymansk Yeah, fresh context

[fofr @fofrAI]

Some tips on how I'm doing discovery work like this:

- make sure your pipeline reproduces known results blindly
- make sure it also rejects known imposters and instrument glitches
- never trust your results, always spin up an independent agent with a non-leading prompt to do an adversarial review, do this repeatedly
- have the reviewer re-verify from raw data
- whenever a reviewer finds something, try to distill a generic lesson from that into a skill or process

[fofr @fofrAI]

I found a weird star with Argon's help. Gaia DR3 4582145718700356096.

It's normal in visible light, stable and similar to our sun. In March 2010 it appeared ~60x brighter than expected in deep infrared heat (22 microns).

As if something around Mars-distance is catching 1.3% of the star's light and radiating it as heat.

Probably a huge cloud of debris from a major planetary collision. Or a radio-silent galaxy hiding behind it. But also the tiniest chance it's a technosignature.

Captured by NASA's WISE telescope, no observations at that wavelength since. Needs more data.

[fofr @fofrAI]

@nikodemszymansk It depends, but usually they have to get the data. Sometimes also put in constraints, to try and prove a claim through an alternate route.

[fofr @fofrAI]

@jconorgrogan Yep, super contrived, exactly as ordered 😅

[fofr @fofrAI]

@TomLikesRobots Yep

[fofr @fofrAI]

I keep seeing contrived "Mate in X" chess riddles on my timeline, so I asked Gemini 4 to make some.

[fofr @fofrAI]

@btr_fs 😅

[fofr @fofrAI]

A nanoputian macromolecule in LaTeX ChemFig, made by Gemini 4.

Pelicanocyclophane-co-velocipedane 1, mass 2485.47 Da and weight 2485.47 g/mol.

[fofr @fofrAI]

@blueemi99 That is cool, nice work.

[fofr @fofrAI]

I asked Argon to add a bullet time 3D rotation, and it added a little flying jump and a fish catch too.

[fofr @fofrAI]

A pelican on a bike as a code art quine. Running the code produces the next frame, and so on. Made by Gemini 4 Argon.

[fofr @fofrAI]

Yep, all are from a 33hr window March 15-17 2010 (WISE ran out of outer coolant in August 2010, before its second pass)

Argon measured the offset against the star's 2010 Gaia position and it is very small, about a tenth of a WISE pixel:

- 22 microns: 0.22 ± 0.15 arcseconds southwest
- 12 microns: 0.25 ± 0.09 arcseconds south

[fofr @fofrAI]

I found a weird star with Argon's help. Gaia DR3 4582145718700356096.

It's normal in visible light, stable and similar to our sun. In March 2010 it appeared ~60x brighter than expected in deep infrared heat (22 microns).

As if something around Mars-distance is catching 1.3% of the star's light and radiating it as heat.

Probably a huge cloud of debris from a major planetary collision. Or a radio-silent galaxy hiding behind it. But also the tiniest chance it's a technosignature.

Captured by NASA's WISE telescope, no observations at that wavelength since. Needs more data.

[fofr @fofrAI]

@knowixbuilds I have some blog posts to write.

[fofr @fofrAI]

@p_rabtsevich The CLI or the IDE?
https://t.co/mjO6XuIq2f

[fofr @fofrAI]

@p_rabtsevich Try it with Antigravity CLI, definitely worth it.

[fofr @fofrAI]

@ravartd same tbh

[fofr @fofrAI]

This was made with SQL.

Argon wrote 217 lines of SQL, no SVGs or bitmaps, to render this pelican.

https://t.co/qTMx2Up0Zp