Benedict Quartey
@Benedict_Q
Son of the living God. Building instruction-following intelligence that improves with experience. PhD @Brownuniversity . Organizer @DeepIndaba . Prev @rai_inst
🚨 What is the best way to use foundation models in robotics? Our new work shows that combining LLMs & VLMs with ideas from formal methods leads to robots that can verifiably follow complex, open-ended instructions in the real world. 🌍 We evaluate on over 150 tasks🚀 🧵 (1/4)
Because intellectuals and entrepreneurs usually optimize different value functions. Obsession vs Outcomes, Elegance vs Scalable Execution. When an intellectual rewires their mind to ship, to commoditize their obsession, that’s when true magic happens.
What prevents Boston from helping smart people do more startups? Why do those people have to move to SF to start a company?
We're running out of sci-fi movies from which to build startups
Hilarious because I have been writing with em dashes in papers since undergrad. Now I have to actively make sure I don’t include them sigh
Coworker just sent me an email that uses two em dashes
Boston is the intellectual capital. Not just of the US but of the world.
I'm impressed by the quality of AI tools we already have available to us just 5 years after the release of GPT-3 But unfortunately, we're still in that phase where the TOOLS are great & our capacity to effectively USE them isn't there yet 100s of millions of ChatGPT users will…
ChatGPT can now do work for you using its own computer. Introducing ChatGPT agent—a unified agentic system combining Operator’s action-taking remote browser, deep research’s web synthesis, and ChatGPT’s conversational strengths.
Still making business decisions in spreadsheets? You’re not alone — 69% of SMBs still run their workflows manually. It’s costing you time, clarity, and growth. Here’s how Papermap AI fixes that
"Nobody is an atheist at 50x leverage" is the realest thing ever written on this app
It’s a good model Sir
🚀 Hello, Kimi K2! Open-Source Agentic Model! 🔹 1T total / 32B active MoE model 🔹 SOTA on SWE Bench Verified, Tau2 & AceBench among open models 🔹Strong in coding and agentic tasks 🐤 Multimodal & thought-mode not supported for now With Kimi K2, advanced agentic intelligence…
Can an AI model predict perfectly and still have a terrible world model? What would that even mean? Our new ICML paper formalizes these questions One result tells the story: A transformer trained on 10M solar systems nails planetary orbits. But it botches gravitational laws 🧵
One incredible thing a PhD helps you build is the knowledge that you can approach any complex problem and with some grit, stubborness and time you will crack it! Regardless of prior experience and the technical complexity.
We interact with dogs through touch -- a simple pat can communicate trust or instruction. Shouldn't interacting with robot dogs be as intuitive? Most commercial robots lack tactile skins. We present UniTac: a method to sense touch using only existing joint sensors! [1/5]
How can 🤖 learn from human workers to provably reduce their workload in factories? Our latest @RoboticsSciSys paper answers this question by proposing the first cost-optimal interactive learning (COIL) algorithm for multi-task collaboration.
Artificial intelligence isn’t THE THING … it is the thing that gets us to THE THING!
I herald his beginning … i herald your end … I herald ….
NEWS🚨: An interstellar object has been spotted entering the solar system at high speed
Who are your BEST customers and how do you find more of them? Alan reverse-engineers your top clients using your existing data. No SQL. No data analyst. Just results.
After weeks of building, we’re launching Papermap AI and with it, your first AI employee: Alan 🧠🚀 This isn’t another dashboard or chatbot. Alan is an intelligent agent that turns vague business questions into data-driven insights and actions. It thinks, researches, writes…
Compelling product and great execution. Congratulations @im_roy_lee @neelyweely23 and the entire team @cluely
introducing @cluely. today is the start of a world where you never have to think again. we just killed 9 industries (thread):
Apt @karpathy ! Not enough credit is given to folks building functional reliable LLM based software. Requires a new set of skills not just traditional engineering. All of a sudden nothing is cleanly deterministic, and you have to manage a temperamental demigod living in your…
+1 for "context engineering" over "prompt engineering". People associate prompts with short task descriptions you'd give an LLM in your day-to-day use. When in every industrial-strength LLM app, context engineering is the delicate art and science of filling the context window…