Hasan Poonawala
@hasanpoonawala.bsky.social
Robots, algorithms, tinkering.
https://poonawalalab.github.io/
https://poonawalalab.github.io/
Loving Justin Carpentier's talk right now. Nice workshop!
We’re ready to roll for Day 2 of the #AI4RoboticsWorkshop !
Livestream starts 9am CET:🎥 tinyurl.com/bdtk2nzs
Today’s great lineup of speakers: @dcremers.bsky.social –
@ericbrachmann.bsky.social - Aniruddha Kembhavi - Adrien Gaidon - Nicolas Mansard & Justin Carpentier
Livestream starts 9am CET:🎥 tinyurl.com/bdtk2nzs
Today’s great lineup of speakers: @dcremers.bsky.social –
@ericbrachmann.bsky.social - Aniruddha Kembhavi - Adrien Gaidon - Nicolas Mansard & Justin Carpentier
November 21, 2025 at 3:52 PM
Loving Justin Carpentier's talk right now. Nice workshop!
Reposted by Hasan Poonawala
Collaborative news curation as a service, demonstrating the value of social media based on protocols, not platforms and a “playbook for killing information voids.”
On Tuesday night, you may have noticed something a little different on Bluesky. Here's a peek behind the scenes of how the election feed came together. Endlessly grateful for this collaboration with @wnyc.org/@gothamist.com & @bsky.app!
About The Other Night...
How Graze is defining the next era of a decentralized, open social attention economy.
graze.leaflet.pub
November 6, 2025 at 10:05 PM
Collaborative news curation as a service, demonstrating the value of social media based on protocols, not platforms and a “playbook for killing information voids.”
Reposted by Hasan Poonawala
your embeddings are not safe!
every prompt directly maps to its embedding and back. they’re isomorphic
SipIt is a linear time algorithm for quickly and efficiently extracting input text from embeddings
www.arxiv.org/abs/2510.15511
every prompt directly maps to its embedding and back. they’re isomorphic
SipIt is a linear time algorithm for quickly and efficiently extracting input text from embeddings
www.arxiv.org/abs/2510.15511
October 28, 2025 at 12:03 PM
your embeddings are not safe!
every prompt directly maps to its embedding and back. they’re isomorphic
SipIt is a linear time algorithm for quickly and efficiently extracting input text from embeddings
www.arxiv.org/abs/2510.15511
every prompt directly maps to its embedding and back. they’re isomorphic
SipIt is a linear time algorithm for quickly and efficiently extracting input text from embeddings
www.arxiv.org/abs/2510.15511
What Typst's blistering speed achieves is the accuracy and control of a compiled document with the instant feedback of a WYSIWYG editor. I used latex and tikz for a decade but I'm never going back.
god, it's going to be such a lift to learn but typst really is an improvement over latex
October 17, 2025 at 12:41 PM
What Typst's blistering speed achieves is the accuracy and control of a compiled document with the instant feedback of a WYSIWYG editor. I used latex and tikz for a decade but I'm never going back.
Reposted by Hasan Poonawala
yesterday i advised self-censorship to a undergrad student im helping apply for grfp re: their research & personal motivations. i think they've got as good a shot as any other. i think im making the right call, but it feels deeply wrong and capitulatory. any thoughts (esp from reviewers)?
October 15, 2025 at 8:32 PM
yesterday i advised self-censorship to a undergrad student im helping apply for grfp re: their research & personal motivations. i think they've got as good a shot as any other. i think im making the right call, but it feels deeply wrong and capitulatory. any thoughts (esp from reviewers)?
Reposted by Hasan Poonawala
I started working on using multi-agent RL to improve self driving in 2021. I started thinking about it in 2018. We're finally seeing this stuff run on cars in 2025. Research is a pretty long-haul thing.
October 14, 2025 at 3:59 PM
I started working on using multi-agent RL to improve self driving in 2021. I started thinking about it in 2018. We're finally seeing this stuff run on cars in 2025. Research is a pretty long-haul thing.
Reposted by Hasan Poonawala
prototype done
official announcement soon
I'm still improving the ratatui firmware (live on yt)
github.com/orhun/tuitar
#rustlang #ratatui #embedded
official announcement soon
I'm still improving the ratatui firmware (live on yt)
github.com/orhun/tuitar
#rustlang #ratatui #embedded
August 8, 2025 at 8:16 PM
prototype done
official announcement soon
I'm still improving the ratatui firmware (live on yt)
github.com/orhun/tuitar
#rustlang #ratatui #embedded
official announcement soon
I'm still improving the ratatui firmware (live on yt)
github.com/orhun/tuitar
#rustlang #ratatui #embedded
We're pushed to classify LLMs as either encyclopedias or alphabet soup. That leaves no agency in reacting to its responses.
Maybe religious books, even priests, are a better analogy for LLMs, or at least how we should interact with them. Our agency and psyche are parts of that interaction.
Maybe religious books, even priests, are a better analogy for LLMs, or at least how we should interact with them. Our agency and psyche are parts of that interaction.
June 26, 2025 at 3:43 PM
We're pushed to classify LLMs as either encyclopedias or alphabet soup. That leaves no agency in reacting to its responses.
Maybe religious books, even priests, are a better analogy for LLMs, or at least how we should interact with them. Our agency and psyche are parts of that interaction.
Maybe religious books, even priests, are a better analogy for LLMs, or at least how we should interact with them. Our agency and psyche are parts of that interaction.
Reposted by Hasan Poonawala
Tyler has an exciting new work that achieves orders-of-magnitude faster asymptotically optimal planning with and without hardware acceleration. This work is under review at IEEE RA-L and an early release will be presented at the RoboARCH workshop at ICRA 2025...
robotic-esp.com/news/2025_05...
robotic-esp.com/news/2025_05...
Asymptotically Optimal RRT-Connect (AORRTC)
Tyler has an exciting new work that achieves orders-of-magnitude faster asymptotically optimal planning with and without hardware acceleration. This work is with Wil Thomason and Zak Kingston (CoMMA L...
robotic-esp.com
May 19, 2025 at 3:25 PM
Tyler has an exciting new work that achieves orders-of-magnitude faster asymptotically optimal planning with and without hardware acceleration. This work is under review at IEEE RA-L and an early release will be presented at the RoboARCH workshop at ICRA 2025...
robotic-esp.com/news/2025_05...
robotic-esp.com/news/2025_05...
Nice post. Teleop is indeed hard, especially when you are dealing with contact. It'll be interesting to follow a company's transition from teleop to autonomous operation.
"That robot video is just teleoperated." Well, teleoperation has its own substantial challenges in robotics. Good, powerful robot teleoperation requires high performance networking and solving a challenging and ambiguous control problem. Blog post: open.substack.com/pub/itcanthi...
Remote Robotic Teleoperation
Far from being an "easy out" for autonomy, teleoperating robots has a range of unique challenges, but enables impressive real-world capabilities nonetheless
open.substack.com
May 18, 2025 at 6:13 PM
Nice post. Teleop is indeed hard, especially when you are dealing with contact. It'll be interesting to follow a company's transition from teleop to autonomous operation.
Reposted by Hasan Poonawala
I pushed out another blog post! How to close a drawer, how to do just about any manipulation, based on techniques described by Dan Whitney in a talk I attended about 50 years ago.
It is only an eight-minute read. Hope you like it.
mtmason.com/the-heart-of...
It is only an eight-minute read. Hope you like it.
mtmason.com/the-heart-of...
The Heart of Robotic Manipulation - Matthew T. Mason
Contact is the heart of robotic manipulation. To understand manipulation, you must understand contact.
mtmason.com
May 6, 2025 at 2:59 PM
I pushed out another blog post! How to close a drawer, how to do just about any manipulation, based on techniques described by Dan Whitney in a talk I attended about 50 years ago.
It is only an eight-minute read. Hope you like it.
mtmason.com/the-heart-of...
It is only an eight-minute read. Hope you like it.
mtmason.com/the-heart-of...
Reposted by Hasan Poonawala
Very interesting oral history -- interviews with some top NLP folks on the effects of GenAI on their field:
www.quantamagazine.org/when-chatgpt...
www.quantamagazine.org/when-chatgpt...
When ChatGPT Broke an Entire Field: An Oral History | Quanta Magazine
Researchers in “natural language processing” tried to tame human language. Then came the transformer.
www.quantamagazine.org
May 1, 2025 at 2:44 AM
Very interesting oral history -- interviews with some top NLP folks on the effects of GenAI on their field:
www.quantamagazine.org/when-chatgpt...
www.quantamagazine.org/when-chatgpt...
Reposted by Hasan Poonawala
Part 2 of SLAM handbook is out for public comments! let us know what you think :-) Issue tracker on GitHub awaits! Link: github.com/SLAM-Handboo...
March 31, 2025 at 8:44 PM
Part 2 of SLAM handbook is out for public comments! let us know what you think :-) Issue tracker on GitHub awaits! Link: github.com/SLAM-Handboo...
Reposted by Hasan Poonawala
About fifty years ago, NASA strapped a message in a bottle to a rocket and flung it into the deep dark
It wasn’t supposed to go this far, but it did. Long past its mission, it’s still out there so far away now that a simple hello takes a day to reach it, and another day to hear if it says hello back
It wasn’t supposed to go this far, but it did. Long past its mission, it’s still out there so far away now that a simple hello takes a day to reach it, and another day to hear if it says hello back
March 12, 2025 at 5:04 AM
About fifty years ago, NASA strapped a message in a bottle to a rocket and flung it into the deep dark
It wasn’t supposed to go this far, but it did. Long past its mission, it’s still out there so far away now that a simple hello takes a day to reach it, and another day to hear if it says hello back
It wasn’t supposed to go this far, but it did. Long past its mission, it’s still out there so far away now that a simple hello takes a day to reach it, and another day to hear if it says hello back
Reposted by Hasan Poonawala
This week's #PaperILike is "Model Predictive Control and Reinforcement Learning: A Unified Framework Based on Dynamic Programming" (Bertsekas 2024).
If you know 1 of {RL, controls} and want to understand the other, this is a good starting point.
PDF: arxiv.org/abs/2406.00592
If you know 1 of {RL, controls} and want to understand the other, this is a good starting point.
PDF: arxiv.org/abs/2406.00592
Model Predictive Control and Reinforcement Learning: A Unified Framework Based on Dynamic Programming
In this paper we describe a new conceptual framework that connects approximate Dynamic Programming (DP), Model Predictive Control (MPC), and Reinforcement Learning (RL). This framework centers around ...
arxiv.org
March 2, 2025 at 4:19 PM
This week's #PaperILike is "Model Predictive Control and Reinforcement Learning: A Unified Framework Based on Dynamic Programming" (Bertsekas 2024).
If you know 1 of {RL, controls} and want to understand the other, this is a good starting point.
PDF: arxiv.org/abs/2406.00592
If you know 1 of {RL, controls} and want to understand the other, this is a good starting point.
PDF: arxiv.org/abs/2406.00592
Reposted by Hasan Poonawala
18F was doing exactly the type of work that DOGE claims to want – yet we were eliminated shortly after midnight. Read our letter to the American people:
18f.org
18f.org
We're not done yet | 18F
18f.org
March 1, 2025 at 11:38 PM
18F was doing exactly the type of work that DOGE claims to want – yet we were eliminated shortly after midnight. Read our letter to the American people:
18f.org
18f.org
Reposted by Hasan Poonawala
2/10 When we do not understand things well and try to engineer with them we are likely to have some very bad outcomes occasionally and to waste of lot of resources and very smart people's lives trying to get them to work right. [[Over 50% of Isaac Newton's working life was lost to alchemy.]]
February 17, 2025 at 4:35 PM
2/10 When we do not understand things well and try to engineer with them we are likely to have some very bad outcomes occasionally and to waste of lot of resources and very smart people's lives trying to get them to work right. [[Over 50% of Isaac Newton's working life was lost to alchemy.]]
Reposted by Hasan Poonawala
Every Jan 1 I post a scorecard on predictions I made, with dates, on Jan 1, 2018 on cars (self-driving), robots, AI, & ML, and on human spaceflight. Besides telling which turned out right and which wrong in the last year I also talk a lot of smack about these topics. rodneybrooks.com/predictions-...
Predictions Scorecard, 2025 January 01 – Rodney Brooks
rodneybrooks.com
January 1, 2025 at 7:37 AM
Every Jan 1 I post a scorecard on predictions I made, with dates, on Jan 1, 2018 on cars (self-driving), robots, AI, & ML, and on human spaceflight. Besides telling which turned out right and which wrong in the last year I also talk a lot of smack about these topics. rodneybrooks.com/predictions-...
Reposted by Hasan Poonawala
If you are a scientist (including graduate students) or professional science communicator who is at least 18 years old and are using Bluesky for any kind of professional purposes, please consider taking our survey.
Link:
umiami.qualtrics.com/jfe/form/SV_...
Please help share. 🧪🌎 #SciComm
Link:
umiami.qualtrics.com/jfe/form/SV_...
Please help share. 🧪🌎 #SciComm
December 30, 2024 at 3:04 PM
If you are a scientist (including graduate students) or professional science communicator who is at least 18 years old and are using Bluesky for any kind of professional purposes, please consider taking our survey.
Link:
umiami.qualtrics.com/jfe/form/SV_...
Please help share. 🧪🌎 #SciComm
Link:
umiami.qualtrics.com/jfe/form/SV_...
Please help share. 🧪🌎 #SciComm
Watched the one on genAI in robotics, really good.
Takeaway: (Software analogy) I think robots will use traditional robotics like C++ exes (solid compiled algorithms under the hood) and genAI like Python glue code (mix and match the solid algorithms at run time). System 2 & 1 if you prefer.
Takeaway: (Software analogy) I think robots will use traditional robotics like C++ exes (solid compiled algorithms under the hood) and genAI like Python glue code (mix and match the solid algorithms at run time). System 2 & 1 if you prefer.
Just discovered the recordings of the debates this year at ICRA. As usual, it was a very popular event. It's easily often the highlight of the conference for me, very cool that they're uploaded! I wish an event of this type was held at every conference.
Robotics Debates
Debates on the Future of Robotics Research
Full-Day Hybrid Workshop | May 13th at ICRA 2024
Time: 10:00-16:35 JST | Pacific Convention Plaza Yokohama, Room G411 + Live streaming here!
www.roboticsdebates.org
December 4, 2024 at 6:26 PM
Watched the one on genAI in robotics, really good.
Takeaway: (Software analogy) I think robots will use traditional robotics like C++ exes (solid compiled algorithms under the hood) and genAI like Python glue code (mix and match the solid algorithms at run time). System 2 & 1 if you prefer.
Takeaway: (Software analogy) I think robots will use traditional robotics like C++ exes (solid compiled algorithms under the hood) and genAI like Python glue code (mix and match the solid algorithms at run time). System 2 & 1 if you prefer.
Reposted by Hasan Poonawala
Wrote a short blog post about my recent experiments with Embassy, the #rustlang framework for writing embedded async applications: pramode.net/2024/12/01/r...
Experiments with Embassy, the embedded async Rust framework
pramode.net
December 1, 2024 at 6:50 PM
Wrote a short blog post about my recent experiments with Embassy, the #rustlang framework for writing embedded async applications: pramode.net/2024/12/01/r...
Reposted by Hasan Poonawala
It is really hard to get robot policies that are both precise (small margins for error) and general (robust to variations).
We just released ViSk, where skin sensing is used to train fine-grained policies with ~1 hour of data. I have attached a single-take video on this post.
visuoskin.github.io
We just released ViSk, where skin sensing is used to train fine-grained policies with ~1 hour of data. I have attached a single-take video on this post.
visuoskin.github.io
October 25, 2024 at 5:57 PM
It is really hard to get robot policies that are both precise (small margins for error) and general (robust to variations).
We just released ViSk, where skin sensing is used to train fine-grained policies with ~1 hour of data. I have attached a single-take video on this post.
visuoskin.github.io
We just released ViSk, where skin sensing is used to train fine-grained policies with ~1 hour of data. I have attached a single-take video on this post.
visuoskin.github.io
Reposted by Hasan Poonawala
Our lab is looking for a postdoc working on multi-robot autonomy, control and perception. This may include work on combining robots with recent advances in preception such as VLMs. If you are interested or are looking for more details about the project, hit up @ericeaton.bsky.social . #Robotics
We're hiring a postdoc in field robotics at the University of Pennsylvania's GRASP lab, which will focus on multi-robot autonomy, control/coordination, and perception for the DARPA Triage Challenge. #robotics
For details and to apply, see www.seas.upenn.edu/~eeaton/open...
For details and to apply, see www.seas.upenn.edu/~eeaton/open...
Open Research Positions
www.seas.upenn.edu
November 28, 2024 at 5:52 PM
Our lab is looking for a postdoc working on multi-robot autonomy, control and perception. This may include work on combining robots with recent advances in preception such as VLMs. If you are interested or are looking for more details about the project, hit up @ericeaton.bsky.social . #Robotics
Reposted by Hasan Poonawala
Egoless Engineering
It turns out misery is a shitty proxy metric.
egoless.engineering
November 21, 2024 at 4:52 PM
Reposted by Hasan Poonawala
Arxiv sharing reminder
pdf ❌
abs ✅
pdf ❌
abs ✅
November 26, 2024 at 8:42 AM
Arxiv sharing reminder
pdf ❌
abs ✅
pdf ❌
abs ✅