{"schema":"OTR-CURRENT-SIGNAL-READER-1","presentationSchema":"OTR-SIGNAL-SLIDES-1","signalId":"SIG-35b513e076b1c1fcce93aad2-10","canonicalUrl":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/","title":"Robots are learning through rehearsal","category":"Technology","representation":"current-guided-reader","originalResearch":"preserved-separately","sources":[{"title":"developers.googleblog.com ↗","url":"https://developers.googleblog.com/gemini-4-argon-for-robotics/","contexts":["What changed","Evidence, memory & review 4 linked sources"]},{"title":"bostondynamics.com ↗","url":"https://bostondynamics.com/blog/robot-hands-for-modern-ai-and-real-work/","contexts":["What changed","Evidence, memory & review 4 linked sources"]},{"title":"www.kaist.ac.kr ↗","url":"https://www.kaist.ac.kr/newsen/html/news/?GotoPage=1&mng_no=67610&mode=V&skey=keyword&sval=Marathon","contexts":["What changed","Evidence, memory & review 4 linked sources"]},{"title":"www.nature.com ↗","url":"https://www.nature.com/articles/s41586-026-11102-5","contexts":["What changed","Evidence, memory & review 4 linked sources"]}],"chapters":[{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-1","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-1","title":"OTR EMERGING DATA","topic":"emerging","kind":"cover","text":"Start with the main idea: what may be changing, how the clues connect, and what remains uncertain. WE see robot progress coming from a useful loop: build a realistic practice world, rehearse many variations, then test the real machine. But better software cannot erase weak grips, short operating time or repair needs. The next convincing advance must report the whole job. Google Developers: Gemini 4 Argon for robotics · Boston Dynamics: Robot hands for modern AI and real work · KAIST: Marathon-capable RAIBO2 research","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-2","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-2","title":"The idea, step by step","topic":"emerging","kind":"research","text":"Start with the main idea: what may be changing, how the clues connect, and what remains uncertain. Read the matching explanation in OTR EMERGING DATA .","links":["https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-1"]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-3","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-3","title":"Research context","topic":"emerging","kind":"research","text":"Understand the background and earlier comparisons, including the differences that matter. Publication and review times First published 22 Oct 2026, 16:41 SGT Research review No fixed date Horizon: 18 months. Review status LEARNING REVIEWED WITH DISCLOSED LIMITS","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-4","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-4","title":"The key clues","topic":"emerging","kind":"research","text":"See the main observations behind the idea, with their important qualifications. Google’s 13 October report describes more than 4,500 simulated training episodes. In one physical workflow, it reports success rising from 56% to 75% and completion time falling from 35 to 24 seconds. These are vendor-reported results for that workflow. Google Developers: Gemini 4 Argon for robotics Published: 13 Oct 2026; publication time not supplied | Observed: Vendor experiment dates and trial denominator not disclosed in retrieved article | Retrieved: 2026-10-22 15:05 SGT Boston Dynamics’ current hand-design account stresses repairability, ruggedness and accurate simulation. KAIST’s separate endurance report shows why energy-aware design also matters, while warning that its research robot is not yet a mass-produced product. Boston Dynamics: Robot hands for modern AI and real work Published: Publication date not displayed in retrieved primary page | Observed: Current manufacturer design description checked on retrieval | Retrieved: 2026-10-22 15:05 SGT KAIST: Marathon-capable RAIBO2 research Published: 28 Sep 2026; publication time not supplied | Observed: Marathon completed in November 2024; journal publication 23 Sep 2026 | Retrieved: 2026-10-22 15:03 SGT","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-5","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-5","title":"What could change this view?","topic":"changes","kind":"research","text":"See what could change this interpretation and what still needs to be checked. WE would weaken the view if simulated practice loses its gains on unfamiliar objects, or if intervention, energy and repair costs outweigh the time saved in a real workday.","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-6","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-6","title":"What this means","topic":"emerging","kind":"research","text":"Read the takeaway and the limits that matter before drawing a conclusion. For workers and small organisations: define a narrow task and its safe failure behaviour before judging a robot demo. Ask who resets it, how often help is needed and how long it can keep working. Those answers matter more than human-like movement. Google Developers: Gemini 4 Argon for robotics · Boston Dynamics: Robot hands for modern AI and real work This is research and interpretation, not an investment price forecast. Source verification limits are listed below.","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-7","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-7","title":"Evidence behind this view","topic":"evidence","kind":"research","text":"Review the information supporting this view and the limits on what it can show. The software and hardware accounts point in the same direction: realistic simulation gets more useful when the physical machine is designed to be modelled, repaired and operated consistently. Neither side alone supplies dependable work. Google Developers: Gemini 4 Argon for robotics Published: 13 Oct 2026; publication time not supplied | Observed: Vendor experiment dates and trial denominator not disclosed in retrieved article | Retrieved: 2026-10-22 15:05 SGT Boston Dynamics: Robot hands for modern AI and real work Published: Publication date not displayed in retrieved primary page | Observed: Current manufacturer design description checked on retrieval | Retrieved: 2026-10-22 15:05 SGT","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-8","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-8","title":"Historical context · What past price moves show","topic":"evidence","kind":"research","text":"Understand the background and earlier comparisons, including the differences that matter. Horizon: 18 months. WE assign 70% subjective confidence that a customer or independent trial will show simulation-trained task gains on unfamiliar objects without increasing human help per task. This is an not shown by testing to give reliable probabilities forecast; it does not predict general home-robot readiness. OTR Database Data earlier connected spatial cameras with cloud-based robot-training data and kept independent training gains unverified. WE now add simulation-to-real results and physical operating constraints. This extends the mechanism; it does not prove the earlier product’s success. OTR Database Data Published: Archive edition 15 Oct 2026 06:00 SGT; underlying company release dated 14 Oct, exact time unknown | Observed: Archive audit 15 Oct 2026 06:13:31 SGT, not exact browser retrieval or a training measurement | Retrieved: 2026-10-22 15:15 SGT Google Developers: Gemini 4 Argon for robotics Published: 13 Oct 2026; publication time not supplied | Observed: Vendor experiment dates and trial denominator not disclosed in retrieved article | Retrieved: 2026-10-22 15:05 SGT KAIST: Marathon-capable RAIBO2 research Published: 28 Sep 2026; publication time not supplied | Observed: Marathon completed in November 2024; journal publication 23 Sep 2026 | Retrieved: 2026-10-22 15:03 SGT","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-9","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-9","title":"Evidence against this view","topic":"changes","kind":"research","text":"Check the opposing evidence and other explanations that could weaken this view. A 75% success rate in one reported workflow still leaves substantial failure. The marathon evidence concerns one research machine and an event from November 2024, not a new October 2026 fleet deployment. Google Developers: Gemini 4 Argon for robotics Published: 13 Oct 2026; publication time not supplied | Observed: Vendor experiment dates and trial denominator not disclosed in retrieved article | Retrieved: 2026-10-22 15:05 SGT KAIST: Marathon-capable RAIBO2 research Published: 28 Sep 2026; publication time not supplied | Observed: Marathon completed in November 2024; journal publication 23 Sep 2026 | Retrieved: 2026-10-22 15:03 SGT","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-10","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-10","title":"How the research changed","topic":"evidence","kind":"research","text":"See what changed in the research and why, while keeping earlier recorded views separate. Learning review before publication The supplied direction was retained: No investible directional call. The supplied synthesis was retained. Reason: Preserve the complete source research, dates, source-reported confidence and contextual horizons. Six actual collaborating Astra agents reviewed all20; primary-source access, historical-corpus and empirical-outcome limits are disclosed. The separate Social03 note clarifies two hypotheses without changing the source wording. No investment direction or forecast contract is issued. No individual outcome lesson was cited in this review. This review belongs to the new draft before publication. Earlier published calls and their original deadlines remain unchanged.","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-11","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-11","title":"What people say vs the facts","topic":"evidence","kind":"research","text":"Compare what people say with the observations, without treating attention as proof. “The robot learned thousands of times” is not the same as “the robot works all day.” WE would look for new-object performance and help rates, not equate simulated episode counts with safe autonomous operation. Google Developers: Gemini 4 Argon for robotics","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-12","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-12","title":"Earlier research and limits","topic":"evidence","kind":"research","text":"Check what earlier forecast results are available and what they can tell us. Past results do not guarantee this outcome. No applied outcome lesson is claimed for this card.","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-13","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-13","title":"The idea step by step","topic":"evidence","kind":"research","text":"Follow the reasoning from the observations to the proposed result. A possible link is not proof. Capture the workplace. Build a simulation. Vary the objects and tasks. Transfer the practice to the robot. Finally, measure failures, human help and energy use in the real setting, not only a polished demonstration.","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-14","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-14","title":"Verification","topic":"evidence","kind":"research","text":"Check what was reviewed, what remains uncertain and how much the sources can support. Sources and verification WE read the original developer report, manufacturer design account and KAIST research report. Nature information describing the data confirms the underlying endurance paper, but its full text was unavailable. The KAIST and Nature records are one study. Google Developers: Gemini 4 Argon for robotics Published: 13 Oct 2026; publication time not supplied | Observed: Vendor experiment dates and trial denominator not disclosed in retrieved article | Retrieved: 2026-10-22 15:05 SGT | Verification: Original developer report read. Results are vendor-reported and task-specific, not independent field validation. Boston Dynamics: Robot hands for modern AI and real work Published: Publication date not displayed in retrieved primary page | Observed: Current manufacturer design description checked on retrieval | Retrieved: 2026-10-22 15:05 SGT | Verification: Original manufacturer design account read. Durability and manufacturability are design priorities, not independently audited fleet outcomes. KAIST: Marathon-capable RAIBO2 research Published: 28 Sep 2026; publication time not supplied | Observed: Marathon completed in November 2024; journal publication 23 Sep 2026 | Retrieved: 2026-10-22 15:03 SGT | Verification: Original institutional report read. One research robot and an older event; multiple-unit reliability and commercial certification remain open. Nature: A quadruped robot designed to complete a marathon on a single battery charge Published: 23 Sep 2026 online publication; exact time not supplied | Observed: Underlying marathon event November 2024 | Retrieved: 2026-10-22 15:16 SGT | Verification: Official publisher information describing the data verified through indexed result. Direct full text was not accessible. Same research as T5-C, not another replication.","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-15","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-15","title":"What’s changing?","topic":"evidence","kind":"research","text":"Follow the reasoning from the observations to the proposed result. A possible link is not proof. The useful shift is from hand-crafting every demonstration toward generating many practice variations. That can make learning cheaper, while pushing the remaining test onto transfer quality and real working hours.","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-16","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-16","title":"Research update","topic":"evidence","kind":"research","text":"Read the details behind this part of the research and how they connect to the main idea. What to check By 22 April 2028, require at least 100 unfamiliar-object attempts against a matched, same-hardware baseline. Success must improve without more human interventions per task; energy per completed job must be reported. No gain or an absent comparison leaves OUR forecast unmet. Google Developers: Gemini 4 Argon for robotics · KAIST: Marathon-capable RAIBO2 research Keep exploring When watching a robot video, ask to see every attempt, the operator’s interventions and the reset time. One successful clip cannot show how much useful work the system completes over a shift.","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-17","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-17","title":"Historical context","topic":"evidence","kind":"research","text":"Understand the background and earlier comparisons, including the differences that matter. Evidence, memory & review OTR Database Data earlier connected spatial cameras with cloud-based robot-training data and kept independent training gains unverified.","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-18","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-18","title":"People and organisations","topic":"evidence","kind":"research","text":"Identify the people, organisations or assets discussed and the part they play in the research. Google supplies the simulation-led workflow. Boston Dynamics describes hardware designed for practical manipulation. KAIST demonstrates energy-aware locomotion. These teams address different pieces; their reports are not a shared end-to-end validation. Google Developers: Gemini 4 Argon for robotics Published: 13 Oct 2026; publication time not supplied | Observed: Vendor experiment dates and trial denominator not disclosed in retrieved article | Retrieved: 2026-10-22 15:05 SGT Boston Dynamics: Robot hands for modern AI and real work Published: Publication date not displayed in retrieved primary page | Observed: Current manufacturer design description checked on retrieval | Retrieved: 2026-10-22 15:05 SGT KAIST: Marathon-capable RAIBO2 research Published: 28 Sep 2026; publication time not supplied | Observed: Marathon completed in November 2024; journal publication 23 Sep 2026 | Retrieved: 2026-10-22 15:03 SGT","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-19","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-19","title":"OTR reasoning · Why we see this connection","topic":"evidence","kind":"research","text":"Follow the reasoning from the observations to the proposed result. A possible link is not proof. WE connect cheap practice with the cost of real failures. Better rehearsal can reduce repeated teaching, but physical wear and recovery still happen outside the simulator. Reliable deployment needs both learning and maintenance evidence. Capture the workplace. Build a simulation.","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-20","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-20","title":"Opposing explanations","topic":"evidence","kind":"research","text":"Check the opposing evidence and other explanations that could weaken this view. Counterpoint A fixed machine or a simple tool may still be the better answer for a repetitive job. General-looking robots earn their place only when flexibility delivers enough useful work to justify the added complexity. Boston Dynamics: Robot hands for modern AI and real work","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-21","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-21","title":"What changed","topic":"evidence","kind":"research","text":"See what changed in the research and why, while keeping earlier recorded views separate. The fresh trigger is the 13 October simulation-to-real workflow report. The hand-design page is undated, and the endurance event is old. WE use them to test the fresh claim, not to make three old events look new. Recorded public history WE see robot progress coming from a useful loop: build a realistic practice world, rehearse many variations, then test the real machine. But better software cannot erase weak grips, short operating time or repair needs. Forecast prepared: 22 Oct, 15:53 SGT Published by OTR: 22 Oct 2026, 16:41 SGT","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-22","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-22","title":"Why the view changed","topic":"evidence","kind":"research","text":"See what changed in the research and why, while keeping earlier recorded views separate. WE add a whole-task test to the earlier data-and-training theme: practice quality matters, but useful output must also survive unfamiliar objects, recovery and the energy budget.","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-23","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-23","title":"Counterpoint · Another possible explanation","topic":"evidence","kind":"research","text":"Check the opposing evidence and other explanations that could weaken this view. Read the matching explanation in Chapter 5 and OTR EMERGING DATA .","links":["https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-5","https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-1"]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-24","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-24","title":"Evidence, memory & review 4 linked sources","topic":"sources","kind":"research","text":"Check where the information came from, when it was observed and what was verified. Source trail This evidence panel does not assign a score or probability. Any forecast estimate is shown separately with its own source and limitations.","links":[]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-25","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-25","title":"Keep exploring","topic":"connections","kind":"finish","text":"Choose how to follow this Signal or compare it with other research. Track this Signal to find it again, compare the evidence, and revisit its recorded outcome. Explore more Signals →","links":["https://ontherice.org/signals/"]},{"id":"otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-26","url":"https://ontherice.org/signals/SIG-35b513e076b1c1fcce93aad2-10/#otr-slide-SIG-35b513e076b1c1fcce93aad2-10-detail-26","title":"Terms in simple English","topic":"glossary","kind":"glossary","text":"Find plain meanings for the specialist terms used in this research. Estimates whose accuracy has not been established by testing A judgement that has not been shown to match real outcomes reliably. It is not a proven success rate. AI or machine learning Software learning patterns from examples to perform tasks. Its output can be wrong. UTC and SGT UTC is a world time standard. SGT is Singapore time, eight hours ahead of UTC. Simulation is a computer practice world. Sim-to-real transfer means learning there and using it on a physical machine. An intervention is a human action needed to help the robot complete or recover a task.","links":[]}]}