TickerTain
TickerTain
NewsroomShortsPortfolioConvergence
NewsroomShortsPortfolioConvergence
$ANTHROPIC·$MA····$INTC····$BLUE-ORIGIN·$SPCX····$CRWV····$CRM····$MSFT····$NVDA····$ORCL····$CURSOR·$AAPL····$OPENAI·$AMZN····$UBER····$GOOGL····$META····$TSLA····$DATABRICKS·$PERPLEXITY·$LYFT····$NBIS····$TSM····$LITE····$ANDURIL·
$ANTHROPIC·$MA····$INTC····$BLUE-ORIGIN·$SPCX····$CRWV····$CRM····$MSFT····$NVDA····$ORCL····$CURSOR·$AAPL····$OPENAI·$AMZN····$UBER····$GOOGL····$META····$TSLA····$DATABRICKS·$PERPLEXITY·$LYFT····$NBIS····$TSM····$LITE····$ANDURIL·
←

chelsea finn

T3 · host / generalist

Chelsea Finn founded Physical Intelligence two years ago to develop general-purpose robot foundation models. She presents research on scalable reinforcement learning, multi-timescale memory, and compositional generalization for robotics. She previously completed a PhD and discusses the tradeoffs between academic research and industry impact.

6 calls·5 names·100% bull·last heard last month·Y Combinator
track record

no scored calls yet — needs a stated position or a categorical verdict, with a matured window vs SPY

top calls

highest conviction · one per company
1sthigh conviction
$PHYSICAL-INTELLIGENCEPhysical Intelligenceposition

Physical Intelligence achieves general-purpose robot policies matching specialist performance

PI's π0.7 model matches or exceeds fine-tuned specialist models across diverse tasks without task-specific training, demonstrating a GPT-like leap for robotics where a single pre-trained policy works out of the box.

Y Combinator2026-08episode →
2ndhigh conviction
$WAYMOWaymo

Waymo's 250k weekly autonomous rides prove trustworthy physical AI is possible

Waymo crossing 250,000 weekly fully autonomous rides demonstrates that ML-based systems can operate reliably in the physical world, providing a proof point for general physical AI.

Y Combinator2026-08episode →
3rdhigh conviction
$GOOGLAlphabet (Waymo)

Waymo's 250k weekly autonomous rides prove trustworthy physical AI autonomy is achievable

Waymo's achievement of 250,000 weekly fully autonomous rides demonstrates that machine learning systems can operate trustworthily and autonomously in the physical world, providing a proof of concept for general physical AI deployment.

Y Combinator2026-08episode →

most discussed · click a bar to filter

  • $PHYSICAL-INTELLIGENCE
  • $WEAVE
  • $WAYMO
  • $ULTRA
  • $GOOGL

recurring themes

  • Robotics & Physical AI6
  • AI Infrastructure2
  • Memory & Storage1
6 total
$PHYSICAL-INTELLIGENCE
Physical Intelligence
HIGHchelsea finn·Y Combinator·last month·Chelsea Finn: This is the State of the Art in Robotics· position
Physical Intelligence achieves general-purpose robot policies matching specialist performance
PI's π0.7 model matches or exceeds fine-tuned specialist models across diverse tasks without task-specific training, demonstrating a GPT-like leap for robotics where a single pre-trained policy works out of the box.
"we see that the across the board the single PIO like pre-trained PIO7 model matches or outperforms the fine-tuned specialists that were developed with reinforcement learning post-…"
30:42
$WEAVE
Weave
MEDchelsea finn·Y Combinator·last month·Chelsea Finn: This is the State of the Art in Robotics
YC company Weave deploys Physical Intelligence models for warehouse packaging
Weave is using PI models for warehouse packaging tasks, demonstrating early commercial adoption of general-purpose robot policies.
"the two videos on the top are actually two YC companies Ultra and Weave uh that have taken find uh PI models and post-trained them to do uh in deployment to do tasks like folding…"
38:16
$WAYMO
Waymo
HIGHchelsea finn·Y Combinator·last month·Chelsea Finn: This is the State of the Art in Robotics
Waymo's 250k weekly autonomous rides prove trustworthy physical AI is possible
Waymo crossing 250,000 weekly fully autonomous rides demonstrates that ML-based systems can operate reliably in the physical world, providing a proof point for general physical AI.
"a year ago Whimo passed the uh quarter of a million weekly autonomous rides suggesting that it is really possible to develop a machine learning based system that can uh operate in…"
5:11
$PHYSICAL-INTELLIGENCE
Physical Intelligence
HIGHchelsea finn·Y Combinator·last month·Chelsea Finn: This is the State of the Art in Robotics
Physical Intelligence demonstrates general-purpose robot model matching specialist performance
Physical Intelligence's PI07 model achieves general-purpose robotic manipulation by training on diverse data with detailed prompting, matching or exceeding fine-tuned specialist models on tasks like laundry folding, box building, and espresso making, while demonstrating compositional generalization to unseen appliances and robot platforms.
"two years ago, I founded a company called physical intelligence. And uh we're really interested in how we can basically uh develop any robot allow any robot to do any task in the…"
0:17
$ULTRA
Ultra
MEDchelsea finn·Y Combinator·last month·Chelsea Finn: This is the State of the Art in Robotics
YC company Ultra deploys Physical Intelligence models for laundry folding in production
Ultra, a Y Combinator company, has successfully post-trained Physical Intelligence's foundation models for real-world deployment on laundry folding tasks, demonstrating commercial viability of general-purpose robotics models.
"the two videos on the top are actually two YC companies Ultra and Weave uh that have taken find uh PI models and post-trained them to do uh in deployment to do tasks like folding…"
38:18
$GOOGL
···
Alphabet (Waymo)
HIGHchelsea finn·Y Combinator·last month·Chelsea Finn: This is the State of the Art in Robotics
Waymo's 250k weekly autonomous rides prove trustworthy physical AI autonomy is achievable
Waymo's achievement of 250,000 weekly fully autonomous rides demonstrates that machine learning systems can operate trustworthily and autonomously in the physical world, providing a proof of concept for general physical AI deployment.
"a year ago Whimo passed the uh quarter of a million weekly autonomous rides suggesting that it is really possible to develop a machine learning based system that can uh operate in…"
5:13
9
Robotics & Physical AItailwind
Scalable RL recipe delivers 90%+ reliability for long-horizon autonomous tasks
Combining human-in-the-loop intervention to avoid dead-end trajectories with a general value function trained on diverse robot experience enables 2x throughput gains and 90%+ success rates on complex manipulation tasks, proving reinforcement learning can achieve the reliability required for real-world deployment.
9
Robotics & Physical AItailwind
General-purpose robot models achieve specialist-level performance out of the box
Physical Intelligence's PI07 model matches or exceeds fine-tuned specialist models across diverse tasks (espresso making, box building, folding) without task-specific training, marking a 'GPT moment' for robotics where a single pre-trained model works out of the box.
9
Robotics & Physical AItailwind
General-purpose robot policies reach GPT-like inflection with π0.7
Physical Intelligence's π0.7 model achieves out-of-the-box performance matching or exceeding task-specific RL-finetuned specialists across diverse manipulation tasks, marking a transition from BERT-era specialist models to GPT-era generalist policies for robotics.
8
AI Infrastructuretailwind
RL post-training with human interventions and learned value functions yields 2x throughput gains
A scalable RL recipe combining early human intervention to avoid dead-end trajectories and a generalizable value function trained on diverse robot experience doubles task throughput and achieves >90% success on complex manipulation like espresso making.
8
Robotics & Physical AItailwind
Multi-timescale memory unlocks 10-15 minute non-repetitive autonomous workflows
A memory architecture combining short-term video memory (10 seconds) with long-term textual summaries (10-15 minutes) enables robots to execute multi-step, non-repetitive tasks like kitchen cleaning autonomously, solving a critical bottleneck for real-world utility.
8
Robotics & Physical AItailwind
Compositional generalization enables zero-shot transfer to new appliances and robot morphologies
π0.7 demonstrates Dolly-like compositional generalization: operating unseen appliances (air fryers) and transferring folding skills to a completely different bimanual robot platform without any morphology-specific training data.
7
Memory & Storagetailwind
Multi-timescale memory (video + text summaries) unlocks 15-minute autonomous horizons
Combining short-term video memory (10s) with long-term compressed text summaries (10-15min) enables robots to execute non-repetitive, multi-step tasks like kitchen cleaning autonomously, solving the context window bottleneck for physical AI.
7
AI Infrastructuretailwind
Diverse data + metadata prompting unlocks low-quality data for generalist training
Ablation shows diverse data is critical for generalization, and prompting with metadata (data quality, episode length) lets models extract signal from low-quality data instead of being hurt by it, enabling internet-scale heterogeneous training.