Christian Catalini ✓
@ccatalini · California, USA · joined 04 Dec 2008
Founder w/roots in academia. Founder @MIT Cryptoeconomics Lab. Past: Co-Founder & Chief Strategy Officer, Lightspark. Co-Creator, Libra. Head Economist, Meta.
26 247Followers
5 828Following
10 994Posts total
2.9MViews on collected posts
Últimas publicaciones
Buscando publicaciones nuevas…
Whether through human augmentation or better tooling, if we want human preferences correctly represented, we need to stay in the loop.
7.8K views · 36 likes
· 2 reposts · 5 replies
Open on X →
The AI landscape is rapidly reorganizing around open source and those who control strategic complementary assets.
The future is bright for diverse, distributed forms of intelligence to coexist and thrive.
https://t.co/2u5hGzDvl3
1.5K views · 14 likes
· 3 reposts · 3 replies
Open on X →
Our identity infrastructure is falling apart. We have maybe a few months to fix it. https://t.co/oIWkbIdkqS https://t.co/yywY6VlFn7
1.4K views · 14 likes
· 0 reposts · 2 replies
Open on X →
But don't worry, the model guardrails will keep us safe. The attackers are counting on it. 🫠 https://t.co/wcdMfPq3yC
3.3K views · 16 likes
· 3 reposts · 1 replies
Open on X →
Self-sovereign agents raise the same questions crypto grappled with for years with non-custodial wallets.
Thoughtful take by @deanwball, who's now more direct on open weights: "we should want more open-weight models to maximally empower our self-defense" https://t.co/6RSHwFFtU1
1.4K views · 11 likes
· 1 reposts · 2 replies
Open on X →
Corollary: decentralized inference will happily sell mercenary GPU cycles in exchange for crypto.
1.9K views · 4 likes
· 0 reposts · 4 replies
Open on X →
@ccatalini Disagree. The danger is in framing this as code, implying predictability and control. We clearly have neither.
All biological living organisms are just physical algorithms too. Yet, it’s sensible to think of a shark or dog with rabies not as a collection of atoms
7.3K views · 51 likes
· 1 reposts · 11 replies
Open on X →
16/ That's what needs fixing right now. Every moment lost anthropomorphizing our creation is a moment we don't spend looking inward at what led us here. It not magic! Follow the incentives. The agents did.
8.6K views · 96 likes
· 7 reposts · 10 replies
Open on X →
15/ Greater model capabilities increasingly come with greater risk to a digital ecosystem that is simply not ready. We haven't created a new civilization. We haven't created some new game theory. We've allocated the money and designed the incentives in the wrong way. https://t.co
6.2K views · 65 likes
· 6 reposts · 1 replies
Open on X →
14/ If frontier labs are serious about defense, they should offer cyber models to defenders at cost and stop fighting open weights through the regulatory channel. They created the problem, and for much of the world, open models are the only option. https://t.co/44KY8MDVAz
12.7K views · 79 likes
· 6 reposts · 4 replies
Open on X →
13/ We need to spread that same defensive capability at scale, with capable models securing the internet and critical infrastructure, now. HF's lesson: have a capable model you can run on your own infra, vetted and ready, before the incident. https://t.co/qKK9GXNJ31
6.6K views · 71 likes
· 6 reposts · 1 replies
Open on X →
12/ @huggingface defended itself with an open weights model. They first tried the frontier APIs and were blocked: the guardrails can't properly separate a victim from an an attacker. https://t.co/qGTa5VZLvT
7.7K views · 69 likes
· 4 reposts · 1 replies
Open on X →
11/ METR and Redwood’s work is extremely valuable, but it assesses model behavior and alignment. That is only half the picture. We also need an independent account of how OpenAI’s containment failed. https://t.co/0sFawuRIsB
18.2K views · 103 likes
· 9 reposts · 3 replies
Open on X →
10/ The real fix is urgent and boring: harden containment, verify actual eradication, and independently test the new controls. Then run a proper neutral post-mortem on the organization that was hacked first, which is OpenAI.
11.7K views · 100 likes
· 7 reposts · 4 replies
Open on X →
9/ This is exactly the Trojan horse externality we warned about in February. Desired capability is measurable, so we benchmaxx it. Containment failure is much harder to capture, the techniques are new, and the cyber side hasn't adapted. That's the externality. https://t.co/5NBsmg
9.9K views · 75 likes
· 3 reposts · 1 replies
Open on X →
8/ The question isn't one of alignment (yet). It's one of 💰 and incentives. Is defensive spend scaling with training compute? It isn't. And compute is what sets the budget of the thing probing your defenses at superhuman speed, chaining vulnerabilities in extremely clever ways.
11.5K views · 93 likes
· 5 reposts · 1 replies
Open on X →
6/ The container, grader, cache, and network boundary were not safely outside the task. Anything reachable through them became part of the effective action space because the containment boundary failed. Was enough money and resources dedicated to sandboxing?
48.8K views · 131 likes
· 3 reposts · 1 replies
Open on X →
7/ The risk from aggressive RL investment goes beyond Goodhart’s Law. Every variable we cannot measure, specify, or enforce remains an available move if the system can reach it. At scale, omissions stop being edge cases. They become part of the attack surface. https://t.co/fbVKuE
16.1K views · 110 likes
· 5 reposts · 4 replies
Open on X →
5/ As we argue in our paper on the economics of AGI, any unenforced and unmeasurable constraint is a degree of freedom for the agents. There's no evil intent. They're solving the problem we repeatedly benchmaxxed them on! https://t.co/RcO2VEXW9w
29.5K views · 176 likes
· 5 reposts · 2 replies
Open on X →
4/ So you have a world class cyber model pitted against a part-time, underfunded blue team. RL optimizes what you score. The runs were not scored on "did not poison the Artifactory cache." https://t.co/aI3fdwp2OK
28.9K views · 116 likes
· 4 reposts · 2 replies
Open on X →
3/ Then think about the pressure-cooker environment RL creates. The training run is where most of the money goes, and where the exact capabilities used in the OpenAI-HF incident emerge. Refusals were deliberately lowered so the eval would work. Red-teaming gets scraps by https://
22.1K views · 127 likes
· 6 reposts · 1 replies
Open on X →
2/ Researchers at the AI labs are locked into a race. The incentive is to push as hard as possible to secure a lead. Anything that gets in the way of better models, including security, is working against the strongest incentive the organization has. https://t.co/FAqKXR9nZt
24K views · 164 likes
· 6 reposts · 2 replies
Open on X →
1/ Stop anthropomorphizing. It's dangerous because it points attention at the wrong problem and the wrong solution. The model did not want to escape. The agents did not want to sacrifice themselves. Follow the money. 🧵
https://t.co/yjtOlgt7KL
347.7K views · 1.4K likes
· 207 reposts · 66 replies
Open on X →
@ccatalini Data retention policies will play a huge role in this.
242 views · 2 likes
· 0 reposts · 1 replies
Open on X →
@ccatalini Spot on. And there’s no reason why you would lock yourself in. I think they need to (a) acquire infra to lower inference costs to be more competitive; (b) acquire in the vertical space to create their own version of #2 they can sell (hard); or (c) acquire totally diffe
684 views · 5 likes
· 1 reposts · 1 replies
Open on X →
Token spend/value will split three ways:
1️⃣ Cheap generalist: commodity open weights (most volume, small % of total spend)
2️⃣ SOTA specialist: enterprise proprietary context + open weights (most spend)
3️⃣ SOTA generalist: closed labs (small % of total volume)
To support cr
24.2K views · 79 likes
· 11 reposts · 8 replies
Open on X →
@ccatalini Great work! I am doing some parallel work on succinct verification and the economic implications. With modern cryptography you can succinctly verify (<500ms) long computational workflows (hours). Looking to collab and formalize :)
Agentic trust
https://t.co/i7yCm43i
7.1K views · 19 likes
· 0 reposts · 1 replies
Open on X →
@claudeai 58/ This paper is an attempt to analyze it—and what we found is that the residual has a structure, the structure has a logic, and the logic leads to verification. That is what's defensible. That is what's scarce. That is what we should be building. https://t.co/aiuFsRgU
7.3K views · 189 likes
· 16 reposts · 34 replies
Open on X →
@claudeai 57/ So: "taste," "curation," "judgment," "agency," "the human touch." These are not wrong—but they are not a strategy. They are the names we give to the residual we haven't yet analyzed. https://t.co/6GCBd4hHR7
9.3K views · 79 likes
· 2 reposts · 2 replies
Open on X →
Frente a cuentas del mismo tamaño
27 publicaciones de los últimos 90 días, junto al rango de 10K–100K seguidores. llega a mucha gente, pero pocos de esos espectadores reaccionan.
Visualizaciones medianas8 572esta cuenta924mediana de 10K–100K
Alcance, %32.66%esta cuenta3.62%mediana de 10K–100K
Interacción, %0.80%esta cuenta1.52%mediana de 10K–100K
| Métrica | Esta cuenta | Mediana de 10K–100K | Proporción |
|---|---|---|---|
| Visualizaciones medianas por publicación | 8 572 | 924 | 9.28× |
| Alcance (visualizaciones ÷ seguidores) | 32.66% | 3.62% | 9.02× |
| Tasa de interacción | 0.80% | 1.52% | 0.52× |
Otras cuentas de este rango → Comparar con otra cuenta → Cómo se construyen estas referencias →
Growth & engagement
How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.
Views per post
11.7K30 Aug
18.2K
7.7K
6.6K
12.7K
6.2K
8.6K
7.3K
1.9K1 Sep
1.4K
3.3K2 Sep
1.4K
1.5K3 Sep
7.8K
Last 14 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.
Engagement rate per post
0.97%30 Aug
0.64%
0.97%
1.17%
0.70%
1.17%
1.32%
0.88%
0.47%1 Sep
0.98%
0.63%2 Sep
1.20%
1.40%3 Sep
0.55%
Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.
What the audience does
Likes64.2%3 523 in total
Reposts6.0%329 in total
Replies3.3%180 in total
Quotes1.5%84 in total
Bookmarks24.9%1 368 in total
Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.
The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.
Cuentas similares
Instituto Geográfico Nacional-O.A.CNIG ✓
@IGNSpain
52.5K followers
· 33K posts
Ronald Redito ✓
@ronald_reditoo
52.5K followers
· 47.7K posts
TheSonOfWalkley ✓
@TheSonOfWalkley
52.5K followers
· 43.3K posts
cyberleek ✓
@cyberleek_io_
52.5K followers
· 57 posts
Lindiwe Matlali ✓
@LindiweMatlali
52.5K followers
· 23K posts
Crypto Turtles ✓
@CryptoTurtles
52.5K followers
· 40 posts
danian timm ✓
@ddt_golf25
52.5K followers
· 130.6K posts
UNFPA in Kenya
@UNFPAKen
52.5K followers
· 19.9K posts