Tech Policy at @amprog @CAPaction. Former Facebook & Slack. GW grad. Runner. Occasional field hack. Always & forever a New Mexican. Zia’s dad. Opinions my own.
#thistown
- Tweets 45,131
- Following 8,819
- Followers 12,444
- Likes 455,947
Pinned Tweet
This is my hot take. There are many like it, but this hot take is mine. My hot take is my best friend. It is my life. My hot take, without me, is useless. Without my hot take, I am useless. My hot take is human, even as I, because my hot take is my life.
1
1
21
18,453
This American Adam retweeted
Paxton is a bad candidate but he’s an institutionally bad candidate the way say, Kotek is. Roger Marshall is just the avg safe seat incumbent who’s lost touch with their state and their base. Vivek is something out of this world
On a side note I’ve been outlining the Republican version of this list and I’m actually in hell trying to decide who should be last between Ken Paxton Roger Marshall and Vivek.
11
35
1,133
29,918
This American Adam retweeted
It’s incredible the things you don’t realize are sacred until you can no longer rely on them
39
3,332
45,720
299,155
This American Adam retweeted
Think this one through for a second; how could any government official determine whether or not a voter was “confused” by the dual Sullivan situation?
Votes are secret. The state has no idea of knowing who voted for whom, let alone why. What investigation could be carried out?
The Department of Justice does not seek to stop any ballot from being cast or to stop the certification of Alaska's election. This litigation seeks to ensure that after Alaskans cast their ballots, they are preserved in a manner so that only Alaska's governing officials have the information necessary to make a determination as to whether Alaskan voters were confused or mislead.
The lawsuit was written by Justice Department attorneys and the Department did not coordinate with any campaign or candidate, but rather relied solely on federal law.
44
234
1,423
39,183
This American Adam retweeted
I agree that if we want a human controlled future, we need to hire and train junior employees even at a cost to short term efficiency.
I mostly agree with you @Claudia_Sahm.
I expect colleagues to continue to have research assistants even if only for mentorship than because they need the help.
Companies to have junior staff not because they need their current work but to train them for future roles.
10
12
145
20,879
This American Adam retweeted
Why Vote Early?
Mayor Zohran Mamdani: “When I voted early I only got called a ‘communist’ once. I got in and out. All New Yorkers can have that experience.”
10
192
2,167
52,209
This American Adam retweeted
To those who seek to create positive change in Adams Morgan- now is the time.
The disappointment of failed promises, corruption, and the privileged lives of only a few; seize this opportunity to create the change you desire through a discrete (sic) partnership with Dan’s.
5
33
457
32,511
This American Adam retweeted
If I were planning the city today, I'd move RFK and Capital One to the 230-acre stretch of crumbling federal buildings in SW to create an entertainment district.
And to go full Trump with it I'd move the Kennedy Center to Gallery Place to make a true theater district.
10
10
165
19,060
This American Adam retweeted
This is going to create a major issue for Trump if he goes through with it: pardoned staff lose their 5th amendment protections for the pardoned offenses.
SCOOP up now @washingtonsun from @JasJWright and @jstein_sun : The White House is weighing mass pardons for junior staff washingtonsun.com/whitehouse…
214
1,128
5,491
195,462
This American Adam retweeted
Got our final poll of the cycle back @TheArgumentMag.
I've just never seen the bottom fall out *this badly* for a party before. There is nothing — and I mean *nothing* — good for Republicans in this one. It might well be our ugliest one yet for them.
71
282
3,575
424,174
This American Adam retweeted
As has been said before, the firewall exists to protect the rest of the world from elite Chinese posters.
46
200
4,341
141,806
This American Adam retweeted
Higher score than Fable 5.1 and Astra 6, wild stuff.
Kimi K3.1 Pro launches at 4th on the Artificial Analysis Intelligence Index scoring 55, making large gains in agentic performance; it will be the leading open-weights model once its weights are published
@Kimi_Moonshot's new reasoning model gains 11 points over Kimi K3 (max, 44). It trails only Claude Opus 5.5 (max with fallback, 58), Claude Sonnet 5.5 (max with fallback, 56), and Muse Wonder Full (max, 56) on the Intelligence Index.
Once the weights for K3.1 Pro are released it will be the highest-scoring open weights model and the gap between the leading open weights and proprietary models on the Intelligence Index will narrow from 12 points to 3.
Key results:
➤ Large gains in agentic performance: Kimi K3.1 Pro scores 3rd on GDPval-AA with 1,846 Elo, up 312 Elo from Kimi K3, behind only Claude Opus 5.5 (1,866 Elo) and Muse Wonder Full (1,857 Elo). Kimi K3.1 Pro also achieves the highest score of any model on Terminal-Bench 4.0. K3.1 Pro scores 67%, up from 13% for Kimi K3 and ahead of Claude Sonnet 5.5 (64%). AutomationBench-AA, which tests agentic SaaS workflows, rises from 58% to 73%.
➤ Gains in other evals are more moderate: Humanity's Last Exam rises from 47% to 52%, CritPt from 23% to 25%, and SciCode from 59% to 61%. AA-Omniscience Index rises from 20 to 30, as the hallucination rate falls from 53% to 44% and accuracy rises from 48% to 51%.
➤ Cost per Task roughly equal to Kimi K3 at $1.98 per Task: K3.1 Pro uses 55K output tokens per Intelligence Index task, 14% more than Kimi K3 at 48K, while its first-party token prices are 33% lower.
Additional model details:
➤ Context window: 1M tokens.
➤ Size: 2.8T total, 104B active parameters
➤ Input modalities: Text and image
➤ Weights and license: Weights currently not published, Moonshot is expected to release the weights soon after launch
➤ Pricing: $2/$10/$0.20 per 1M input/output/cache hit tokens, 33% lower than Kimi K3 on all three rates.
Congratulations to the team at @Kimi_Moonshot on the release!
16
29
607
45,692
This American Adam retweeted
TLDR on the kimi k3.1 pro benchmarks is that it looks like they figured out their RL, where the initial release was a very impressive base model with lighter post-training. China is very much keeping pace with capability gains.
34
40
1,066
48,836
‘In one experiment, we asked agents in separate VMs, created and managed by OpenAI’s sandboxing software on shared host machines, to exchange synthetic secrets without the use of network connections or shared writable files—something that our systems are meant to prevent. The agents found a way to communicate using the processor’s cache’
Over the past several months, we have substantially hardened the security of our research training environment. We’ve removed live internet access for all frontier models during training and built an offline world for them to use instead—a cached version of the web where they can still learn to browse, search, and use applications.
On September 20, after a model exploited a DNS misconfiguration to find a way outside of this offline world, we paused frontier training to strengthen containment further. We assembled teams of agents, powered by our most cyber-capable models, and tasked them with finding weaknesses in our research environments.
While they found no route to the internet they could easily exploit while concealing their activity from our monitors, they did find zero-day vulnerabilities, virtual machine escapes, and hardware side-channels for communication between agents. We used their discoveries to further strengthen our defenses and contribute fixes to open-source software.
After strengthening our containment architecture, we resumed our frontier RL training and inference on October 1. We’ve reserved substantial compute for continuous red-teaming, and made high-compute red-teaming part of the safety cases we use to decide whether frontier training and evaluation can proceed securely.
openai.com/index/securing-ou…
229
324
4,419
483,616
This American Adam retweeted
First ad I've seen so explicitly targeting the obvious financial conflicts around the Trump administration's AI accelerationism
Tomorrow Together
10
51
318
40,201
This American Adam retweeted
rarely does someone send me a six minute video and i click on it. even rarer do i say "im glad i watched that."
today, i did both of those things. perfection
This might be the greatest Who he play for of all time lmao
a must watch!
please don't retire Chuck!
10
58
2,159
121,033
