Josep Gisbert
@josep-gisbert.bsky.social
Assistant Professor of Financial Economics @IEUniversity. Personal Finance and AI 🫶 Researching: Macro-Finance, Banking, AI, and Behavioral Economics.
To learn more, visit: https://www.josepgisbert.com
To learn more, visit: https://www.josepgisbert.com
📄 Paper: esade.edu/ecpol/wp-con...
#GenAI #Inequality #RCT #AIinEducation #IEUniversity #ToniRoldán #EsadeEcPol #EconomicsResearch
#GenAI #Inequality #RCT #AIinEducation #IEUniversity #ToniRoldán #EsadeEcPol #EconomicsResearch
May 12, 2025 at 1:01 PM
• 🛠️ 𝐔𝐬𝐚𝐠𝐞 𝐠𝐚𝐩 𝐦𝐚𝐭𝐭𝐞𝐫𝐬: The real differentiator was not access to GenAI, but how well students could prompt, interpret, and apply its output
𝐀 𝐟𝐫𝐞𝐬𝐡 𝐩𝐞𝐫𝐬𝐩𝐞𝐜𝐭𝐢𝐯𝐞 𝐨𝐧 𝐡𝐨𝐰 𝐆𝐞𝐧𝐀𝐈 𝐜𝐨𝐮𝐥𝐝 𝐚𝐜𝐭𝐮𝐚𝐥𝐥𝐲 𝐰𝐢𝐝𝐞𝐧, 𝐫𝐚𝐭𝐡𝐞𝐫 𝐭𝐡𝐚𝐧 𝐧𝐚𝐫𝐫𝐨𝐰, 𝐬𝐤𝐢𝐥𝐥 𝐠𝐚𝐩𝐬 𝐞𝐬𝐩𝐞𝐜𝐢𝐚𝐥𝐥𝐲 𝐢𝐧 𝐬𝐞𝐭𝐭𝐢𝐧𝐠𝐬 𝐭𝐡𝐚𝐭 𝐝𝐞𝐦𝐚𝐧𝐝 𝐣𝐮𝐝𝐠𝐦𝐞𝐧𝐭 𝐧𝐮𝐚𝐧𝐜𝐞 𝐚𝐧𝐝 𝐫𝐞𝐚𝐥-𝐭𝐢𝐦𝐞 𝐭𝐡𝐢𝐧𝐤𝐢𝐧𝐠
𝐀 𝐟𝐫𝐞𝐬𝐡 𝐩𝐞𝐫𝐬𝐩𝐞𝐜𝐭𝐢𝐯𝐞 𝐨𝐧 𝐡𝐨𝐰 𝐆𝐞𝐧𝐀𝐈 𝐜𝐨𝐮𝐥𝐝 𝐚𝐜𝐭𝐮𝐚𝐥𝐥𝐲 𝐰𝐢𝐝𝐞𝐧, 𝐫𝐚𝐭𝐡𝐞𝐫 𝐭𝐡𝐚𝐧 𝐧𝐚𝐫𝐫𝐨𝐰, 𝐬𝐤𝐢𝐥𝐥 𝐠𝐚𝐩𝐬 𝐞𝐬𝐩𝐞𝐜𝐢𝐚𝐥𝐥𝐲 𝐢𝐧 𝐬𝐞𝐭𝐭𝐢𝐧𝐠𝐬 𝐭𝐡𝐚𝐭 𝐝𝐞𝐦𝐚𝐧𝐝 𝐣𝐮𝐝𝐠𝐦𝐞𝐧𝐭 𝐧𝐮𝐚𝐧𝐜𝐞 𝐚𝐧𝐝 𝐫𝐞𝐚𝐥-𝐭𝐢𝐦𝐞 𝐭𝐡𝐢𝐧𝐤𝐢𝐧𝐠
May 12, 2025 at 1:01 PM
• 🛠️ 𝐔𝐬𝐚𝐠𝐞 𝐠𝐚𝐩 𝐦𝐚𝐭𝐭𝐞𝐫𝐬: The real differentiator was not access to GenAI, but how well students could prompt, interpret, and apply its output
𝐀 𝐟𝐫𝐞𝐬𝐡 𝐩𝐞𝐫𝐬𝐩𝐞𝐜𝐭𝐢𝐯𝐞 𝐨𝐧 𝐡𝐨𝐰 𝐆𝐞𝐧𝐀𝐈 𝐜𝐨𝐮𝐥𝐝 𝐚𝐜𝐭𝐮𝐚𝐥𝐥𝐲 𝐰𝐢𝐝𝐞𝐧, 𝐫𝐚𝐭𝐡𝐞𝐫 𝐭𝐡𝐚𝐧 𝐧𝐚𝐫𝐫𝐨𝐰, 𝐬𝐤𝐢𝐥𝐥 𝐠𝐚𝐩𝐬 𝐞𝐬𝐩𝐞𝐜𝐢𝐚𝐥𝐥𝐲 𝐢𝐧 𝐬𝐞𝐭𝐭𝐢𝐧𝐠𝐬 𝐭𝐡𝐚𝐭 𝐝𝐞𝐦𝐚𝐧𝐝 𝐣𝐮𝐝𝐠𝐦𝐞𝐧𝐭 𝐧𝐮𝐚𝐧𝐜𝐞 𝐚𝐧𝐝 𝐫𝐞𝐚𝐥-𝐭𝐢𝐦𝐞 𝐭𝐡𝐢𝐧𝐤𝐢𝐧𝐠
𝐀 𝐟𝐫𝐞𝐬𝐡 𝐩𝐞𝐫𝐬𝐩𝐞𝐜𝐭𝐢𝐯𝐞 𝐨𝐧 𝐡𝐨𝐰 𝐆𝐞𝐧𝐀𝐈 𝐜𝐨𝐮𝐥𝐝 𝐚𝐜𝐭𝐮𝐚𝐥𝐥𝐲 𝐰𝐢𝐝𝐞𝐧, 𝐫𝐚𝐭𝐡𝐞𝐫 𝐭𝐡𝐚𝐧 𝐧𝐚𝐫𝐫𝐨𝐰, 𝐬𝐤𝐢𝐥𝐥 𝐠𝐚𝐩𝐬 𝐞𝐬𝐩𝐞𝐜𝐢𝐚𝐥𝐥𝐲 𝐢𝐧 𝐬𝐞𝐭𝐭𝐢𝐧𝐠𝐬 𝐭𝐡𝐚𝐭 𝐝𝐞𝐦𝐚𝐧𝐝 𝐣𝐮𝐝𝐠𝐦𝐞𝐧𝐭 𝐧𝐮𝐚𝐧𝐜𝐞 𝐚𝐧𝐝 𝐫𝐞𝐚𝐥-𝐭𝐢𝐦𝐞 𝐭𝐡𝐢𝐧𝐤𝐢𝐧𝐠
🔍 Some standout insights from his exciting 𝐑𝐂𝐓 approach:
• 🧠 𝐇𝐢𝐠𝐡-𝐬𝐤𝐢𝐥𝐥𝐞𝐝 𝐬𝐭𝐮𝐝𝐞𝐧𝐭𝐬 𝐛𝐞𝐧𝐞𝐟𝐢𝐭 𝐦𝐨𝐫𝐞: GenAI improved debate performance far more for merit scholarship recipients and top baseline performers
• ⚖️ 𝐍𝐨𝐭 𝐚 𝐥𝐞𝐯𝐞𝐥𝐢𝐧𝐠 𝐭𝐨𝐨𝐥: Contrary to prior studies, ChatGPT did not help low-performers catch up
• 🧠 𝐇𝐢𝐠𝐡-𝐬𝐤𝐢𝐥𝐥𝐞𝐝 𝐬𝐭𝐮𝐝𝐞𝐧𝐭𝐬 𝐛𝐞𝐧𝐞𝐟𝐢𝐭 𝐦𝐨𝐫𝐞: GenAI improved debate performance far more for merit scholarship recipients and top baseline performers
• ⚖️ 𝐍𝐨𝐭 𝐚 𝐥𝐞𝐯𝐞𝐥𝐢𝐧𝐠 𝐭𝐨𝐨𝐥: Contrary to prior studies, ChatGPT did not help low-performers catch up
May 12, 2025 at 1:01 PM
🔍 Some standout insights from his exciting 𝐑𝐂𝐓 approach:
• 🧠 𝐇𝐢𝐠𝐡-𝐬𝐤𝐢𝐥𝐥𝐞𝐝 𝐬𝐭𝐮𝐝𝐞𝐧𝐭𝐬 𝐛𝐞𝐧𝐞𝐟𝐢𝐭 𝐦𝐨𝐫𝐞: GenAI improved debate performance far more for merit scholarship recipients and top baseline performers
• ⚖️ 𝐍𝐨𝐭 𝐚 𝐥𝐞𝐯𝐞𝐥𝐢𝐧𝐠 𝐭𝐨𝐨𝐥: Contrary to prior studies, ChatGPT did not help low-performers catch up
• 🧠 𝐇𝐢𝐠𝐡-𝐬𝐤𝐢𝐥𝐥𝐞𝐝 𝐬𝐭𝐮𝐝𝐞𝐧𝐭𝐬 𝐛𝐞𝐧𝐞𝐟𝐢𝐭 𝐦𝐨𝐫𝐞: GenAI improved debate performance far more for merit scholarship recipients and top baseline performers
• ⚖️ 𝐍𝐨𝐭 𝐚 𝐥𝐞𝐯𝐞𝐥𝐢𝐧𝐠 𝐭𝐨𝐨𝐥: Contrary to prior studies, ChatGPT did not help low-performers catch up
He presented “𝘞𝘩𝘦𝘯 𝘎𝘦𝘯𝘈𝘐 𝘐𝘯𝘤𝘳𝘦𝘢𝘴𝘦𝘴 𝘐𝘯𝘦𝘲𝘶𝘢𝘭𝘪𝘵𝘺: 𝘌𝘷𝘪𝘥𝘦𝘯𝘤𝘦 𝘧𝘳𝘰𝘮 𝘢 𝘜𝘯𝘪𝘷𝘦𝘳𝘴𝘪𝘵𝘺 𝘋𝘦𝘣𝘢𝘵𝘪𝘯𝘨 𝘊𝘰𝘮𝘱𝘦𝘵𝘪𝘵𝘪𝘰𝘯”, a novel paper that uses a randomized controlled trial (RCT) in an academic debate setting to ask a critical question: Who really benefits from GenAI in cognitively demanding, unpredictable tasks? 📚
May 12, 2025 at 1:01 PM
He presented “𝘞𝘩𝘦𝘯 𝘎𝘦𝘯𝘈𝘐 𝘐𝘯𝘤𝘳𝘦𝘢𝘴𝘦𝘴 𝘐𝘯𝘦𝘲𝘶𝘢𝘭𝘪𝘵𝘺: 𝘌𝘷𝘪𝘥𝘦𝘯𝘤𝘦 𝘧𝘳𝘰𝘮 𝘢 𝘜𝘯𝘪𝘷𝘦𝘳𝘴𝘪𝘵𝘺 𝘋𝘦𝘣𝘢𝘵𝘪𝘯𝘨 𝘊𝘰𝘮𝘱𝘦𝘵𝘪𝘵𝘪𝘰𝘯”, a novel paper that uses a randomized controlled trial (RCT) in an academic debate setting to ask a critical question: Who really benefits from GenAI in cognitively demanding, unpredictable tasks? 📚
• 📉 𝐄𝐟𝐟𝐢𝐜𝐢𝐞𝐧𝐜𝐲: this leads to weaker market efficiency and lower liquidity compared to rational agents
𝐀 𝐭𝐡𝐨𝐮𝐠𝐡𝐭-𝐩𝐫𝐨𝐯𝐨𝐤𝐢𝐧𝐠 𝐜𝐨𝐧𝐭𝐫𝐢𝐛𝐮𝐭𝐢𝐨𝐧 𝐭𝐨 𝐭𝐡𝐞 𝐞𝐦𝐞𝐫𝐠𝐢𝐧𝐠 𝐟𝐢𝐞𝐥𝐝 𝐨𝐟 𝐚𝐥𝐠𝐨𝐫𝐢𝐭𝐡𝐦𝐢𝐜 𝐛𝐞𝐡𝐚𝐯𝐢𝐨𝐫𝐚𝐥 𝐟𝐢𝐧𝐚𝐧𝐜𝐞.
📄 Slides: francescosangiorgi.com/wp-content/u...
#AIinFinance #ReinforcementLearning #MarketDesign
𝐀 𝐭𝐡𝐨𝐮𝐠𝐡𝐭-𝐩𝐫𝐨𝐯𝐨𝐤𝐢𝐧𝐠 𝐜𝐨𝐧𝐭𝐫𝐢𝐛𝐮𝐭𝐢𝐨𝐧 𝐭𝐨 𝐭𝐡𝐞 𝐞𝐦𝐞𝐫𝐠𝐢𝐧𝐠 𝐟𝐢𝐞𝐥𝐝 𝐨𝐟 𝐚𝐥𝐠𝐨𝐫𝐢𝐭𝐡𝐦𝐢𝐜 𝐛𝐞𝐡𝐚𝐯𝐢𝐨𝐫𝐚𝐥 𝐟𝐢𝐧𝐚𝐧𝐜𝐞.
📄 Slides: francescosangiorgi.com/wp-content/u...
#AIinFinance #ReinforcementLearning #MarketDesign
May 9, 2025 at 10:59 AM
• 📉 𝐄𝐟𝐟𝐢𝐜𝐢𝐞𝐧𝐜𝐲: this leads to weaker market efficiency and lower liquidity compared to rational agents
𝐀 𝐭𝐡𝐨𝐮𝐠𝐡𝐭-𝐩𝐫𝐨𝐯𝐨𝐤𝐢𝐧𝐠 𝐜𝐨𝐧𝐭𝐫𝐢𝐛𝐮𝐭𝐢𝐨𝐧 𝐭𝐨 𝐭𝐡𝐞 𝐞𝐦𝐞𝐫𝐠𝐢𝐧𝐠 𝐟𝐢𝐞𝐥𝐝 𝐨𝐟 𝐚𝐥𝐠𝐨𝐫𝐢𝐭𝐡𝐦𝐢𝐜 𝐛𝐞𝐡𝐚𝐯𝐢𝐨𝐫𝐚𝐥 𝐟𝐢𝐧𝐚𝐧𝐜𝐞.
📄 Slides: francescosangiorgi.com/wp-content/u...
#AIinFinance #ReinforcementLearning #MarketDesign
𝐀 𝐭𝐡𝐨𝐮𝐠𝐡𝐭-𝐩𝐫𝐨𝐯𝐨𝐤𝐢𝐧𝐠 𝐜𝐨𝐧𝐭𝐫𝐢𝐛𝐮𝐭𝐢𝐨𝐧 𝐭𝐨 𝐭𝐡𝐞 𝐞𝐦𝐞𝐫𝐠𝐢𝐧𝐠 𝐟𝐢𝐞𝐥𝐝 𝐨𝐟 𝐚𝐥𝐠𝐨𝐫𝐢𝐭𝐡𝐦𝐢𝐜 𝐛𝐞𝐡𝐚𝐯𝐢𝐨𝐫𝐚𝐥 𝐟𝐢𝐧𝐚𝐧𝐜𝐞.
📄 Slides: francescosangiorgi.com/wp-content/u...
#AIinFinance #ReinforcementLearning #MarketDesign
🔍 Some standout insights from their 𝐭𝐡𝐞𝐨𝐫𝐲 + 𝐬𝐢𝐦𝐮𝐥𝐚𝐭𝐢𝐨𝐧:
• ✅ 𝐏𝐫𝐢𝐜𝐞𝐬: AI traders decode prices and learn portfolio strategies that qualitatively mirror rational benchmarks
• ❌ 𝐄𝐱𝐭𝐞𝐫𝐧𝐚𝐥𝐢𝐭𝐢𝐞𝐬: but when multiple AIs interact, they interfere with one another’s learning creating a negative externality
• ✅ 𝐏𝐫𝐢𝐜𝐞𝐬: AI traders decode prices and learn portfolio strategies that qualitatively mirror rational benchmarks
• ❌ 𝐄𝐱𝐭𝐞𝐫𝐧𝐚𝐥𝐢𝐭𝐢𝐞𝐬: but when multiple AIs interact, they interfere with one another’s learning creating a negative externality
May 9, 2025 at 10:59 AM
🔍 Some standout insights from their 𝐭𝐡𝐞𝐨𝐫𝐲 + 𝐬𝐢𝐦𝐮𝐥𝐚𝐭𝐢𝐨𝐧:
• ✅ 𝐏𝐫𝐢𝐜𝐞𝐬: AI traders decode prices and learn portfolio strategies that qualitatively mirror rational benchmarks
• ❌ 𝐄𝐱𝐭𝐞𝐫𝐧𝐚𝐥𝐢𝐭𝐢𝐞𝐬: but when multiple AIs interact, they interfere with one another’s learning creating a negative externality
• ✅ 𝐏𝐫𝐢𝐜𝐞𝐬: AI traders decode prices and learn portfolio strategies that qualitatively mirror rational benchmarks
• ❌ 𝐄𝐱𝐭𝐞𝐫𝐧𝐚𝐥𝐢𝐭𝐢𝐞𝐬: but when multiple AIs interact, they interfere with one another’s learning creating a negative externality
He presented “(𝘋𝘦𝘦𝘱) 𝘓𝘦𝘢𝘳𝘯𝘪𝘯𝘨 𝘵𝘰 𝘛𝘳𝘢𝘥𝘦”, co-authored with Ivan Gufler (@UniLUISS) and Emanuele Tarantino (@UniLUISS & European Commission). The work digs into how AI traders—using deep reinforcement learning—respond to price signals, interact with each other, and influence market outcomes 📚
May 9, 2025 at 10:59 AM
He presented “(𝘋𝘦𝘦𝘱) 𝘓𝘦𝘢𝘳𝘯𝘪𝘯𝘨 𝘵𝘰 𝘛𝘳𝘢𝘥𝘦”, co-authored with Ivan Gufler (@UniLUISS) and Emanuele Tarantino (@UniLUISS & European Commission). The work digs into how AI traders—using deep reinforcement learning—respond to price signals, interact with each other, and influence market outcomes 📚
8.𝐌𝐮𝐥𝐭𝐢-𝐚𝐠𝐞𝐧𝐭? Use a manager or task handoffs, but don’t overcomplicate.
9.𝐆𝐮𝐚𝐫𝐝𝐫𝐚𝐢𝐥𝐬 𝐚𝐫𝐞 𝐧𝐨𝐧-𝐧𝐞𝐠𝐨𝐭𝐢𝐚𝐛𝐥𝐞: protect against bad input, risky actions, and misuse.
10.𝐃𝐞𝐬𝐢𝐠𝐧 𝐟𝐨𝐫 𝐡𝐮𝐦𝐚𝐧𝐬: test early, expect failure, always have a backup plan.
#AI #OpenAI #LLMs #Agents #Product #TechTrends #FutureOfWork
9.𝐆𝐮𝐚𝐫𝐝𝐫𝐚𝐢𝐥𝐬 𝐚𝐫𝐞 𝐧𝐨𝐧-𝐧𝐞𝐠𝐨𝐭𝐢𝐚𝐛𝐥𝐞: protect against bad input, risky actions, and misuse.
10.𝐃𝐞𝐬𝐢𝐠𝐧 𝐟𝐨𝐫 𝐡𝐮𝐦𝐚𝐧𝐬: test early, expect failure, always have a backup plan.
#AI #OpenAI #LLMs #Agents #Product #TechTrends #FutureOfWork
April 19, 2025 at 7:55 PM
8.𝐌𝐮𝐥𝐭𝐢-𝐚𝐠𝐞𝐧𝐭? Use a manager or task handoffs, but don’t overcomplicate.
9.𝐆𝐮𝐚𝐫𝐝𝐫𝐚𝐢𝐥𝐬 𝐚𝐫𝐞 𝐧𝐨𝐧-𝐧𝐞𝐠𝐨𝐭𝐢𝐚𝐛𝐥𝐞: protect against bad input, risky actions, and misuse.
10.𝐃𝐞𝐬𝐢𝐠𝐧 𝐟𝐨𝐫 𝐡𝐮𝐦𝐚𝐧𝐬: test early, expect failure, always have a backup plan.
#AI #OpenAI #LLMs #Agents #Product #TechTrends #FutureOfWork
9.𝐆𝐮𝐚𝐫𝐝𝐫𝐚𝐢𝐥𝐬 𝐚𝐫𝐞 𝐧𝐨𝐧-𝐧𝐞𝐠𝐨𝐭𝐢𝐚𝐛𝐥𝐞: protect against bad input, risky actions, and misuse.
10.𝐃𝐞𝐬𝐢𝐠𝐧 𝐟𝐨𝐫 𝐡𝐮𝐦𝐚𝐧𝐬: test early, expect failure, always have a backup plan.
#AI #OpenAI #LLMs #Agents #Product #TechTrends #FutureOfWork
4.𝐏𝐢𝐜𝐤 𝐭𝐡𝐞 𝐫𝐢𝐠𝐡𝐭 𝐦𝐨𝐝𝐞𝐥: test, explore with GPT-4, then optimize for speed/cost.
5.𝐓𝐨𝐨𝐥𝐬 𝐦𝐚𝐭𝐭𝐞𝐫: agents need reliable tools to access data and take action.
6.𝐈𝐧𝐬𝐭𝐫𝐮𝐜𝐭𝐢𝐨𝐧𝐬 𝐚𝐫𝐞 𝐞𝐯𝐞𝐫𝐲𝐭𝐡𝐢𝐧𝐠: vague = chaos. Be clear and specific.
7.𝐒𝐭𝐚𝐫𝐭 𝐬𝐦𝐚𝐥𝐥 — one solid agent with tools often beats a whole swarm.
5.𝐓𝐨𝐨𝐥𝐬 𝐦𝐚𝐭𝐭𝐞𝐫: agents need reliable tools to access data and take action.
6.𝐈𝐧𝐬𝐭𝐫𝐮𝐜𝐭𝐢𝐨𝐧𝐬 𝐚𝐫𝐞 𝐞𝐯𝐞𝐫𝐲𝐭𝐡𝐢𝐧𝐠: vague = chaos. Be clear and specific.
7.𝐒𝐭𝐚𝐫𝐭 𝐬𝐦𝐚𝐥𝐥 — one solid agent with tools often beats a whole swarm.
April 19, 2025 at 7:55 PM
4.𝐏𝐢𝐜𝐤 𝐭𝐡𝐞 𝐫𝐢𝐠𝐡𝐭 𝐦𝐨𝐝𝐞𝐥: test, explore with GPT-4, then optimize for speed/cost.
5.𝐓𝐨𝐨𝐥𝐬 𝐦𝐚𝐭𝐭𝐞𝐫: agents need reliable tools to access data and take action.
6.𝐈𝐧𝐬𝐭𝐫𝐮𝐜𝐭𝐢𝐨𝐧𝐬 𝐚𝐫𝐞 𝐞𝐯𝐞𝐫𝐲𝐭𝐡𝐢𝐧𝐠: vague = chaos. Be clear and specific.
7.𝐒𝐭𝐚𝐫𝐭 𝐬𝐦𝐚𝐥𝐥 — one solid agent with tools often beats a whole swarm.
5.𝐓𝐨𝐨𝐥𝐬 𝐦𝐚𝐭𝐭𝐞𝐫: agents need reliable tools to access data and take action.
6.𝐈𝐧𝐬𝐭𝐫𝐮𝐜𝐭𝐢𝐨𝐧𝐬 𝐚𝐫𝐞 𝐞𝐯𝐞𝐫𝐲𝐭𝐡𝐢𝐧𝐠: vague = chaos. Be clear and specific.
7.𝐒𝐭𝐚𝐫𝐭 𝐬𝐦𝐚𝐥𝐥 — one solid agent with tools often beats a whole swarm.
Here are 10 key takeaways compiled by the great Hesam @Hesamation:
1.𝐀𝐠𝐞𝐧𝐭𝐬 ≠ 𝐰𝐨𝐫𝐤𝐟𝐥𝐨𝐰𝐬: agents act autonomously; most LLM apps need an input.
2.𝐔𝐬𝐞 𝐭𝐡𝐞𝐦 𝐟𝐨𝐫 𝐭𝐡𝐞 𝐡𝐚𝐫𝐝 𝐬𝐭𝐮𝐟𝐟: messy data, dynamic rules, real decisions.
3.𝐂𝐨𝐫𝐞 𝐫𝐞𝐜𝐢𝐩𝐞 = brain (model), tools, and clear instructions.
1.𝐀𝐠𝐞𝐧𝐭𝐬 ≠ 𝐰𝐨𝐫𝐤𝐟𝐥𝐨𝐰𝐬: agents act autonomously; most LLM apps need an input.
2.𝐔𝐬𝐞 𝐭𝐡𝐞𝐦 𝐟𝐨𝐫 𝐭𝐡𝐞 𝐡𝐚𝐫𝐝 𝐬𝐭𝐮𝐟𝐟: messy data, dynamic rules, real decisions.
3.𝐂𝐨𝐫𝐞 𝐫𝐞𝐜𝐢𝐩𝐞 = brain (model), tools, and clear instructions.
April 19, 2025 at 7:55 PM
Here are 10 key takeaways compiled by the great Hesam @Hesamation:
1.𝐀𝐠𝐞𝐧𝐭𝐬 ≠ 𝐰𝐨𝐫𝐤𝐟𝐥𝐨𝐰𝐬: agents act autonomously; most LLM apps need an input.
2.𝐔𝐬𝐞 𝐭𝐡𝐞𝐦 𝐟𝐨𝐫 𝐭𝐡𝐞 𝐡𝐚𝐫𝐝 𝐬𝐭𝐮𝐟𝐟: messy data, dynamic rules, real decisions.
3.𝐂𝐨𝐫𝐞 𝐫𝐞𝐜𝐢𝐩𝐞 = brain (model), tools, and clear instructions.
1.𝐀𝐠𝐞𝐧𝐭𝐬 ≠ 𝐰𝐨𝐫𝐤𝐟𝐥𝐨𝐰𝐬: agents act autonomously; most LLM apps need an input.
2.𝐔𝐬𝐞 𝐭𝐡𝐞𝐦 𝐟𝐨𝐫 𝐭𝐡𝐞 𝐡𝐚𝐫𝐝 𝐬𝐭𝐮𝐟𝐟: messy data, dynamic rules, real decisions.
3.𝐂𝐨𝐫𝐞 𝐫𝐞𝐜𝐢𝐩𝐞 = brain (model), tools, and clear instructions.
And if you want the TL;DR, the always insightful Greg Isenberg put together a great summary of OpenAI’s best tips — ideal for product builders and growth hackers:
🔗 x.com/gregisenberg...
#PromptEngineering #AI #OpenAI #Google #LLMs #ProductDesign #GenerativeAI
🔗 x.com/gregisenberg...
#PromptEngineering #AI #OpenAI #Google #LLMs #ProductDesign #GenerativeAI
We’ve detected that JavaScript is disabled in this browser. Please enable JavaScript or switch to a supported browser to continue using x.com. You can see a list of supported browsers in our Help Center.
x.com
April 16, 2025 at 12:12 PM
And if you want the TL;DR, the always insightful Greg Isenberg put together a great summary of OpenAI’s best tips — ideal for product builders and growth hackers:
🔗 x.com/gregisenberg...
#PromptEngineering #AI #OpenAI #Google #LLMs #ProductDesign #GenerativeAI
🔗 x.com/gregisenberg...
#PromptEngineering #AI #OpenAI #Google #LLMs #ProductDesign #GenerativeAI
OpenAI meanwhile, shared a practical, no-fluff prompting guide tailored for working with their API — short, sharp, and very useful:
🔗 cookbook.openai.com/examples/gpt...
🔗 cookbook.openai.com/examples/gpt...
GPT-4.1 Prompting Guide | OpenAI Cookbook
Open-source examples and guides for building with the OpenAI API. Browse a collection of snippets, advanced techniques and walkthroughs. Share your own examples and guides.
cookbook.openai.com
April 16, 2025 at 12:12 PM
OpenAI meanwhile, shared a practical, no-fluff prompting guide tailored for working with their API — short, sharp, and very useful:
🔗 cookbook.openai.com/examples/gpt...
🔗 cookbook.openai.com/examples/gpt...
Google released a 69-page whitepaper that dives deep into prompt design for a wide range of applications:
🔗 www.kaggle.com/whitepaper-p...
🔗 www.kaggle.com/whitepaper-p...
April 16, 2025 at 12:12 PM
Google released a 69-page whitepaper that dives deep into prompt design for a wide range of applications:
🔗 www.kaggle.com/whitepaper-p...
🔗 www.kaggle.com/whitepaper-p...
Study highlight: Peer-sourced data outshines official audits in swaying firms vs . Read more on procurement & corruption: loiaconofrancesco.com/wp-content/u...
#DevelopmentEconomics
#DevelopmentEconomics
April 1, 2025 at 5:15 PM
Study highlight: Peer-sourced data outshines official audits in swaying firms vs . Read more on procurement & corruption: loiaconofrancesco.com/wp-content/u...
#DevelopmentEconomics
#DevelopmentEconomics
Key insights from Loiacono’s Uganda field experiment:
• Tender info alone doesn’t lift firm participation
• Fixing misperceptions about public sector integrity boosts bids & contracts
• Tender info alone doesn’t lift firm participation
• Fixing misperceptions about public sector integrity boosts bids & contracts
April 1, 2025 at 5:15 PM
Key insights from Loiacono’s Uganda field experiment:
• Tender info alone doesn’t lift firm participation
• Fixing misperceptions about public sector integrity boosts bids & contracts
• Tender info alone doesn’t lift firm participation
• Fixing misperceptions about public sector integrity boosts bids & contracts
A special thanks to Xavier Vives for his outstanding leadership in organizing this conference, as well as Jordi Canals and the entire 𝐈𝐄𝐒𝐄 𝐁𝐚𝐧𝐤𝐢𝐧𝐠 𝐈𝐧𝐢𝐭𝐢𝐚𝐭𝐢𝐯𝐞 team for making this event possible. Engaging with top academics and industry leaders on the future of AI in finance was truly inspiring.
March 21, 2025 at 1:48 PM
A special thanks to Xavier Vives for his outstanding leadership in organizing this conference, as well as Jordi Canals and the entire 𝐈𝐄𝐒𝐄 𝐁𝐚𝐧𝐤𝐢𝐧𝐠 𝐈𝐧𝐢𝐭𝐢𝐚𝐭𝐢𝐯𝐞 team for making this event possible. Engaging with top academics and industry leaders on the future of AI in finance was truly inspiring.
🔹 AI’s Impact on Financial Markets: The Consequences of Evolving Information Production
• Thierry Foucault, HEC Paris
• Thierry Foucault, HEC Paris
March 21, 2025 at 1:48 PM
🔹 AI’s Impact on Financial Markets: The Consequences of Evolving Information Production
• Thierry Foucault, HEC Paris
• Thierry Foucault, HEC Paris
🔹 𝐂𝐨𝐫𝐩𝐨𝐫𝐚𝐭𝐞 𝐅𝐢𝐧𝐚𝐧𝐜𝐞 𝐚𝐧𝐝 𝐆𝐨𝐯𝐞𝐫𝐧𝐚𝐧𝐜𝐞 𝐰𝐢𝐭𝐡 𝐀𝐈: 𝐎𝐥𝐝 𝐚𝐧𝐝 𝐍𝐞𝐰
• 𝐖𝐞𝐢 𝐉𝐢𝐚𝐧𝐠, Emory University
• 𝐖𝐞𝐢 𝐉𝐢𝐚𝐧𝐠, Emory University
March 21, 2025 at 1:48 PM
🔹 𝐂𝐨𝐫𝐩𝐨𝐫𝐚𝐭𝐞 𝐅𝐢𝐧𝐚𝐧𝐜𝐞 𝐚𝐧𝐝 𝐆𝐨𝐯𝐞𝐫𝐧𝐚𝐧𝐜𝐞 𝐰𝐢𝐭𝐡 𝐀𝐈: 𝐎𝐥𝐝 𝐚𝐧𝐝 𝐍𝐞𝐰
• 𝐖𝐞𝐢 𝐉𝐢𝐚𝐧𝐠, Emory University
• 𝐖𝐞𝐢 𝐉𝐢𝐚𝐧𝐠, Emory University
Then the conference brought together leading academic experts who shared their latest research and perspectives:
🔹 𝐀𝐈 𝐚𝐧𝐝 𝐭𝐡𝐞 𝐅𝐢𝐧𝐚𝐧𝐜𝐢𝐚𝐥 𝐒𝐞𝐜𝐭𝐨𝐫: 𝐓𝐫𝐚𝐧𝐬𝐟𝐨𝐫𝐦𝐚𝐭𝐢𝐨𝐧𝐬, 𝐂𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐞𝐬, 𝐚𝐧𝐝 𝐑𝐞𝐠𝐮𝐥𝐚𝐭𝐨𝐫𝐲 𝐑𝐞𝐬𝐩𝐨𝐧𝐬𝐞𝐬
• Leonardo Gambacorta, Bank for International Settlements
🔹 𝐀𝐈 𝐚𝐧𝐝 𝐭𝐡𝐞 𝐅𝐢𝐧𝐚𝐧𝐜𝐢𝐚𝐥 𝐒𝐞𝐜𝐭𝐨𝐫: 𝐓𝐫𝐚𝐧𝐬𝐟𝐨𝐫𝐦𝐚𝐭𝐢𝐨𝐧𝐬, 𝐂𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐞𝐬, 𝐚𝐧𝐝 𝐑𝐞𝐠𝐮𝐥𝐚𝐭𝐨𝐫𝐲 𝐑𝐞𝐬𝐩𝐨𝐧𝐬𝐞𝐬
• Leonardo Gambacorta, Bank for International Settlements
March 21, 2025 at 1:48 PM
Then the conference brought together leading academic experts who shared their latest research and perspectives:
🔹 𝐀𝐈 𝐚𝐧𝐝 𝐭𝐡𝐞 𝐅𝐢𝐧𝐚𝐧𝐜𝐢𝐚𝐥 𝐒𝐞𝐜𝐭𝐨𝐫: 𝐓𝐫𝐚𝐧𝐬𝐟𝐨𝐫𝐦𝐚𝐭𝐢𝐨𝐧𝐬, 𝐂𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐞𝐬, 𝐚𝐧𝐝 𝐑𝐞𝐠𝐮𝐥𝐚𝐭𝐨𝐫𝐲 𝐑𝐞𝐬𝐩𝐨𝐧𝐬𝐞𝐬
• Leonardo Gambacorta, Bank for International Settlements
🔹 𝐀𝐈 𝐚𝐧𝐝 𝐭𝐡𝐞 𝐅𝐢𝐧𝐚𝐧𝐜𝐢𝐚𝐥 𝐒𝐞𝐜𝐭𝐨𝐫: 𝐓𝐫𝐚𝐧𝐬𝐟𝐨𝐫𝐦𝐚𝐭𝐢𝐨𝐧𝐬, 𝐂𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐞𝐬, 𝐚𝐧𝐝 𝐑𝐞𝐠𝐮𝐥𝐚𝐭𝐨𝐫𝐲 𝐑𝐞𝐬𝐩𝐨𝐧𝐬𝐞𝐬
• Leonardo Gambacorta, Bank for International Settlements