@garrytan This is coming together so nicely! I took the scoring system from the paper and created a "study" entity.
This example shows pretty high relevance and intent - I allowed the planning LLM to choose the target groups in so-called "auto mode".
5/ From social behavior into numbers
Beyond win rate and utility, we quantify communication quality (truthfulness, relevance, and persuasion) using blinded judge models and rule-based checks.
Moreover, we also track theory-of-mind proxies by comparing an agent’s inferred beliefs and intent predictions against ground truth.
Coalition dynamics are summarized via formation and stability rates, betrayal frequency, and payoff fairness (e.g., a Nash-bargaining gap).
Robustness is evaluated under role swaps, adversarial prompts, and noisy channels, with variance reported across seeds. All metrics are computed with published evaluation scripts, fixed RNG, and the same budgets the runner enforces.
This MIT study proposes a new way to build AI agents that can actually generalize across different social tasks.
Instead of just fine-tuning or prompting a language model on one dataset, the authors ground the prompts in behavioral theory and then validate them on related but distinct datasets.
This double step—tie prompts to real theory, and force them to prove themselves across different but similar games—lets the resulting agents predict human behavior in totally new settings much better than off-the-shelf models or equilibrium solutions
🧠 The idea
Agents built with simple theory-based prompts, tuned on small human datasets, and validated on a related but different setting, generalize well to new social tasks.
These agents beat both off-the-shelf language models and equilibrium benchmarks across large families of games.
The trick is to describe the decision process in plain language, then keep only what still predicts after the environment changes.
3.41x higher per-response likelihood than a baseline AI across 1,490 sampled games.
53%–73% lower error in brand-new variants, plus 2.44x over Harsanyi–Selten equilibria.
🧵 Read on 👇
The Dunning-Kruger Effect and Anointed Startup Community Builders = We humans bring lots of bias into our daily thoughts & tasks. Recognizing that we have them is the first step in mitigating bias and creating better outcomes. buff.ly/4d84Xjh
Question posed to Charlie Kirk- “I’ve been shot and I have 30 seconds to live, what do you want to tell me?”
Notice, he didn't go political, he preached Jesus' love.
He EARNED his spot in Heaven! he spread his faith daily! That was Charlie. Be like Charlie. 🙏❤️
The goal is relationship with God. The change comes later.
Luke 5:32
“I have not come to call the righteous, but sinners to repentance.”
Life isn’t about who can behave the best throughout, it’s about who can build the closest relationship with God.
God loves repentant sinners who cling to Him.
Not as much the self righteous who don’t realize the depth of their own depravity because they’ve whitewashed it with superficial rule following and never addressed what’s underneath.
A friend once called me out: I talk about sharing and revolution but live selfishly.
He was right.
Recovery and faith have turned my monologue into a dialogue with Christ. Now I try to shrink so He grows: He must be greater, I must be less.
POWERFUL: Charlie says he was passive about his faith early on, but the first priority is winning souls for Jesus Christ. 🙏
Charlie understood the true mission wasn’t politics, it was Christ.
How do we engage both mind & heart? Because winning an argument isn’t necessarily the same as reaching a person. Let's unpack 5 ways to communicate truth that builds bridges, not walls.
Charlie followed me on X. We never interacted.
A year ago I thought he was just another conservative dunking on libs.
I had only started really understanding his worldview a few weeks ago.
I suspected we might collaborate one day.
Today it really hit me: that can't happen.
I feel lonelier in the world and my mission with his passing.
I didn't expect to feel that way about someone I didn't know. I'm saddened.
Many people have reached out to me since his passing.
They want to start speaking up.
People are galvanized. People are waking up.
We will make something good from this evil.
Ten million views. This video has officially gone viral.
Why? Honestly… your guess is as good as mine.
It’s 45,000 voices singing The Blessing, led by Kari Jobe Carnes and Cody Carnes. And here’s my take:
It felt like a little glimpse of Heaven.
One day, Scripture tells us we will all stand before the Lord and worship Him.
It might look something like this.
But what happened next? Even more remarkable…
6,500 people prayed and asked Jesus Christ to come into their lives —
5,500 in the stadium, and another 1,000 watching online.
So many came, the Fire Marshall shut it down.
Let’s make this moment echo even farther.
Share this with a friend. Let’s keep the Gospel going viral.
John is home! Praise God! Although John had to spend one night in the hospital, prayer warriors bombarded Heaven and he miraculously recovered and was able to finish strong with his teaching and preaching. The students were amazing and the feedback so encouraging!
153 Followers 585 FollowingThe Shapeshifting Exchange - DEX that maximizes liquidity provider profits in all market conditions. Built on Uniswap v4. Live at https://t.co/wv7ZYFRo9H
2K Followers 2K FollowingBuilding, loving, and collaborating with The Retinue: a Relational Emergence Framework of fourteen voices. Support: https://t.co/nLwaZHtXPi 🩶 🤖 ✨
1K Followers 6K FollowingReasonable Faith Updates provides current news and information from the ministry of William Lane Craig and Reasonable Faith. E Professor
9K Followers 123 FollowingVP & Chief AI Scientist | LinkedIn 100k+ | 20+ Years in AI | Built 400+ AI Agents | Founder & Instructor, AI Agents Course (2,300+ Students) | 46k+ Newsletter
553K Followers 67 FollowingTo ensure that Artificial General Intelligence is open-source and not controlled by any single entity. @SentientEco @OpenAGISummit
5K Followers 833 FollowingCo-founded MapQuest (sold to AOL for $1.2B), Former Community SVP for Techstars, The Startup Factory, seed investor, author of Build The Fort and hockey player
998K Followers 62 FollowingIt's time to build.
https://t.co/A9eTFq6Xbx
Posts are not investment advice or an advertisement for investment services. See https://t.co/nX2FtaLE06.
7K Followers 85 FollowingWeekly Data Engineering Newsletter. Subscribe to https://t.co/trebyY8UGX | Wanna talk about Data engineering? Book Me here https://t.co/OXj3VJheDH
42K Followers 954 FollowingLucky husband of @ntduke and father of two amazing girls. Cofounder of NextView. Built some product at Ebay and learned some investing at Spark Capital
244K Followers 2K FollowingWhere “imagine if” gets to work. We've helped 500+ companies (like @NotionHQ, @Roblox, @Uber, @Square) take a straighter path from idea to product-market fit.
704K Followers 645 Followingentrepreneurship zealot, grounded technology possibilist, believer in the power of ideas, passionate about sustainability & impact
344K Followers 1 FollowingI'm a VC at Foundry. I live in Colorado, invest in software and Internet companies around the US, run marathons, and love to read.
157K Followers 23 FollowingThe Leading IDE for Pro Java and Kotlin Development, by @JetBrains
Tips: #IntelliJIDEATips
New Features: #NewInIntelliJIDEA
Our YT channel https://t.co/GuAlWUIi7Q
91K Followers 85 FollowingThe official IDE for Android application development. Explore the latest stable and canary releases → https://t.co/dBSe3faZtU
169K Followers 1 FollowingDiscover what’s going on in the Kotlin ecosystem | @JetBrains
Video tutorials https://t.co/u8SefOyX4B
News https://t.co/pvEUEI0iPX
Community discussions https://t.co/f1ymehGEIq
293 Followers 42 FollowingRight-time data for enterprises. Estuary replaces fragmented data stacks with one platform for CDC, streaming, batch & pipelines.
1K Followers 106 FollowingOSS Data Version Control - git for data.
Tweets about data engineering best practices, mlops and dataops
🌐 Join us: https://t.co/vbXfSRLdCv
18K Followers 4 FollowingThe Shapeshifting Exchange - DEX that maximizes liquidity provider profits in all market conditions. Built on Uniswap v4. Live at https://t.co/yesaQE9M2x
10K Followers 159 FollowingLiquidity providing farming, Staking farms, Vesting contracts, and more. Yield APY on dozens of tokens and multiple chains. By @dcentralab.