min{power}
Home
Notes
Robotics
Computing
Research
Archive
About
Subscribe
Sign in
Latest
Top
Discussions
What an Alternate History of the RISC vs. CISC Debate Teaches Us About Robotics
Explorations in computing and robotics focused on power-efficiency and safety -- personal posts by Avik De, robotics Ph.D. and founder
Jun 18
•
Avik De
and
Bharath Suresh
9
3
Action Tokens: Fine-Grained vs. Behavioral Primitives
Why the LLM tokenization debate matters for physical AI, and what biology tells us
Jun 3
•
Avik De
11
3
May 2026
The Loops and Hierarchies of Embodied Intelligence
Can we get embodied intelligence by connecting cameras and motors to an AI brain?
May 27
•
Avik De
10
2
4
A Multi-Robot Brain is not like a Multi-Chip ISA
"Cross-embodiment" trained policies generalize well, but is that the best solution?
May 15
•
Avik De
11
2
5
How an LLM Changes its Mind
Safety and efficiency with universal approximators and Turing machines
May 5
•
Avik De
6
2
2
April 2026
The First Paradigm in Robotics & AI Research: Lessons from Computer Engineering
Commoditization and end-to-end learning have consolidated robotics and AI. What's next for research labs?
Apr 29
•
Avik De
and
Chris Paxton
32
8
6
Honor's humanoid ran the fastest half-marathon: how did they do it?
Engineering isn't magic, it's a matter of tradeoffs
Apr 22
•
Avik De
8
4
7
Building a reasoning hierarchical robotics pipeline from scratch
Part 5: A demo combining the best features of end-to-end and classical approaches
Apr 7
•
Avik De
9
1
5
March 2026
The Art of Architectural Analysis: Utilization, Throughput, Latency
Putting TinyXPU Under the Microscope
Published on Chip Insights
•
Mar 30
A coding agent equivalent for robotics pipelines
Part 4: Closing the action loop with a VLA vs. a spatial VLM "agent"
Mar 26
•
Avik De
14
3
Lessons from AVs on safety in end-to-end pipelines
Recent developments in autonomous vehicles on recognizing and handling distribution shift
Mar 20
•
Avik De
5
2
1
Mapping Algorithms to Custom Silicon - Part 2
Running matrix multiplication on a TPU-style architecture
Published on Chip Insights
•
Mar 15
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts