Gabriel P. Andrade ✓

@gab_p_andrade

Researcher @GensynAI. Working on game theory, alg econ, multi-agent systems, and decentralized learning. Opinions are my own.

698Followers

213Following

291Posts total

22.2KViews on collected posts

Against accounts of the same size

11 posts from the last 90 days, next to the under 10K follower range. reaches fewer people than peers, but engages them much harder.

Median views145this account5 065median for under 10K

Reach, %20.77%this account349.30%median for under 10K

Engagement, %2.96%this account1.10%median for under 10K

Metric	This account	Median for under 10K	Ratio
Median views per post	145	5 065	0.03×
Reach (views ÷ followers)	20.77%	3.5× audience	0.06×
Engagement rate	2.96%	1.10%	2.69×

Others in this range → Compare with another account → How these benchmarks are built →

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

20.8K8 Jul

248

219

145

120

102

98

93

92

169

146

Last 11 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.12%8 Jul

1.61%

1.83%

2.76%

3.33%

3.92%

4.08%

5.38%

4.35%

2.96%

2.05%

Reactions — likes, reposts, replies and quotes — divided by views. Median for under 10K accounts is 1.10%.

What the audience does

Likes59.5%47 in total

Reposts5.1%4 in total

Replies15.2%12 in total

Quotes3.8%3 in total

Bookmarks16.5%13 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.

Latest posts

11/ Read more about this ongoing work here --> https://t.co/IKAoDIhCh6 We've characterized the 🥔 vs. UMA dichotomy, but have a lot more to say about Oracle mechanisms. We'll follow with a full formal treatment, broader scope, etc. in an extended paper. 146 views · 3 likes · 0 reposts · 0 replies 08 Jul 2026 10/ Importantly, this isn't a blockchain-specific problem. Any automated judge sitting between a question and a consequence (e.g., LLM-as-judge benchmarks, resume screens, moderation appeals) faces the same pressures. 169 views · 4 likes · 0 reposts · 1 replies 08 Jul 2026 9/ As always, there's no free lunch here: LLM errors aren't uniform. Their biases produce concentrated, learnable error patterns that an ex-ante attacker can chase. These error profiles become an audit surface (open weights, bias bounties, etc.) 92 views · 3 likes · 0 reposts · 1 replies 08 Jul 2026 8/ None of this requires AI (recall the 🥔). But an LLM with committed weights, a fixed prompt, and reproducible inference approximates credible neutrality...while being accurate enough on well-specified questions to make the threshold a real barrier. 93 views · 4 likes · 0 reposts · 1 replies 08 Jul 2026 7/ Make the report credibly neutral with error rate ε and the picture flips. Optimal overturn threshold becomes 1−ε. At ε = 5%, an attacker needs a 95% supermajority to overturn a correct report! 98 views · 3 likes · 0 reposts · 1 replies 08 Jul 2026 6/ Why can't UMA just raise the overturn threshold? If reports can be shaped by whoever holds positions, you must guard against every possible report. The best you can achieve is equally guarding in all directions --- an attacker just needs a 50% majority to overturn in their 102 views · 3 likes · 0 reposts · 1 replies 08 Jul 2026 5/ Neutrality without accuracy is a potato. Accuracy without neutrality is a UMA proposer. Both have the same manipulation resistance. A good report layer is simultaneously accurate AND neutral, because then the dispute layer can safely demand a supermajority to overturn. 120 views · 3 likes · 0 reposts · 1 replies 08 Jul 2026 4/ Enter the humble 🥔. A potato is useless at predicting rain. But its output can't be moved by positions, volume, or prices (no matter who bets on what). It satisfies credible neutrality. UMA's report layer doesn't have that; proposers see the market before they propose. 145 views · 3 likes · 0 reposts · 1 replies 08 Jul 2026 3/ A tension surfaces. If we: - centralize the oracle -> single point of failure - naively decentralize it -> the people voting have money riding on the outcome Neither is good, and we've been burnt by both (e.g., WSJ found nine wallets control ~1/2 of UMA voting power on 219 views · 3 likes · 0 reposts · 1 replies 08 Jul 2026 2/ Oracle failures are getting expensive: - $7M settled on zero evidence (one whale held >25% of UMA tokens) - $240M market overturned over whether an outfit counted as a "suit" - A Paris temperature market settled off a single, allegedly hairdryer-tampered airport sensor 248 views · 3 likes · 0 reposts · 1 replies 08 Jul 2026 Do you prefer markets resolved by UMA or a potato? Personally, I want something better than both... 🧵👇 20.8K views · 15 likes · 4 reposts · 3 replies 08 Jul 2026

Similar accounts