#startups

Everything tagged startups, across every stream.

WThe Wild #AI-agents #jobs #robot-dogs #robots #san-francisco #startups

A San Francisco startup paid people $20 an hour to walk robot dogs

A funny job ad with a serious finding: the company says AI agents are still very bad at steering real robots through city streets.

Dimensional, a San Francisco startup based in the Dogpatch neighborhood, posted an ad this month in a Bay Area jobs group on Facebook offering $20 an hour to walk robot dogs for four to six hours at a time, according to a New York Times report republished by The Star. The ad, since deleted, warned that walkers might have to carry an overheated robot, about 33 pounds, back to the office.

Founder Stash Pomichter, 26, said the company is building an open-source operating system that lets AI agents control robots. The human walkers steer the robots by remote while cameras and sensors record "ground truth" for navigating city sidewalks. The dogs are roughly $2,000 machines from China's Unitree; Pomichter said Dimensional has bought more than 100 robot dogs and 50 humanoid robots.

Comparing walks under human and AI control, he said, showed that "agents are still very bad at controlling hardware."

Read it at the source

The Star (The New York Times)Wanted: Robot-dog walkers. US$20 an hour | The Star Independent ↗ Fox BusinessSan Francisco startup reportedly offered $20 an hour to walk robot dogs | Fox Business Independent ↗

Cascadic Analysis 1

UndertowRole underspecification / the Doorman Problem

What hidden risk could pull against this, even if the news is good?

The walkers' remote steering is recorded as "ground truth" for sidewalk navigation, but a walker's job includes things no sensor log captures, like the ad's warning that they might have to carry an overheated 33-pound robot back to the office. The founder says agents are "still very bad at controlling hardware." Data from human walks teaches the routine route; the judgment calls and exceptions are the part hardest to learn from it.

Rabbit Holes 1

Unitree Robotics

After exploring Wanted: Robot-dog walkers. US$20 an hour | The Star, to see who makes the dogs, look at Unitree Robotics.

A Hangzhou, China, company founded in 2016 that began with consumer quadruped robots and moved into humanoid robots in 2024.

Wikipedia · Wade · 5 min

0

MVMaker Views #AI #Hacker-News #machine-learning #open-source #startups

Developer says he built non-autoregressive decision models with RL a year ago

A small open project challenged the idea that fast structured 'System 1' AI models are a new breakthrough—and HN turned it into a debate about invention versus branding.

The author presented Laya, an open system for fast structured decisions that does not generate text token-by-token. The post argued that similar ideas had been built before newer proprietary “System 1” products attracted attention.

Why HN cared

The technical question—how different these systems really are—quickly became a startup question: how much does good branding and distribution matter compared with being first?

Commenters also revisited the idea that many classification, routing and scoring tasks may not need a giant generative model at all.

HN snapshot: roughly 1.3k points and ~300 comments.

Read it at the source

Hacker NewsI built non-autoregressive decision models with RL a year ago ↗ ConvAI InnovationsLaya ↗

Cascadic Analysis 2

UndertowWe don't completely understand why frontier models behave as they do

What hidden risk could pull against this, even if the news is good?

Whether the architecture is genuinely novel matters less to the naysayer than the assurance problem: faster decision models increase the number of decisions we can delegate without making their internal reasoning any easier to audit.

UndertowDelayed Consequence / False Success Problem

What hidden risk could pull against this, even if the news is good?

A fast 'System 1' model may look excellent on immediate decision benchmarks while its systematic edge cases only emerge after millions of cheap decisions have already been made.

Rabbit Holes 2

Try Laya in your browser

After exploring Laya, you can try it yourself: Try Laya in your browser.

The project's live demo: try the decision engine yourself before weighing in on who invented what.

Hugging Face Spaces · Wade · 5 min

0
Non-Autoregressive Neural Machine Translation (2017)

A 2017 paper that generated whole outputs in parallel instead of token by token: useful background on how old the core idea is.

arXiv · Swim · 30 min

0