A local NumPy tracker estimates one upright orange cube from rendered RGB-D and declared calibration. Sixteen held-out pose groups produce 320 correlated fixtures across two resolutions and ten conditions. At 320×240, all 16 clear scenes return estimates, with 1.98 mm position-error p95, 6.36° yaw-error p95 and 7.58 ms estimator wall-time p95. At 160×120, coverage and yaw gates fail. All 256 negative fixtures abstain; partial occlusion returns no estimates in 32 cases and remains unsupported. Independent readback verifies 2,578 files and recomputes all metrics without running the estimator or simulator. No action-learning export or physical control is enabled.
320 fixtures · 16/16 clear estimates at 320 px · no learned skill or hardware qualification
· Frozen four-controller comparison informed by independent design review
A fixed eight-scene CPU comparison tests one bounded incremental pose-hold candidate against three unchanged controls. All 32 attempts are valid. The candidate admits four grasps and terminally refuses each after 320 ms when its proposed cumulative correction exceeds 0.03 rad. One native five-second stationary-grip result includes fallback; no variant achieves a five-second stationary grip with continued controller qualification. Four lower-friction scenes lack admission under every variant. EGM-inspired fresh-context review catches a joint-margin clipping edge case before execution; Harness² inspires a bounded propose/test/retain-lessons workflow. No upstream framework is executed, model trained or controller promoted.
32/32 valid · 0/8 active stationary holds per controller · simulation only
· Three authored hold controllers with separate native grip and TCP-pose endpoints
A frozen twelve-scene CPU study compares measured-joint, issued-joint and bounded Cartesian hold requests without increasing the issued jaw target. All 36 attempts are valid, retaining 720,000 native 1 ms rows and 36,000 commands. Each variant admits four scenes with no false initial qualification; eight lower-friction scenes lack admission under all three. Five-second native grip retention is 2/12, 4/12 and 4/12 respectively, but grip plus common TCP stationarity is 2/12, 0/12 and 0/12. Issued targets retain grips while moving the tool beyond 4 mm. Cartesian terminally refuses all four admitted holds after a position-bound failure; its later native grip retention includes the fallback trajectory. No controller is promoted. This finite authored-control comparison establishes no learned skill, placement improvement, hardware performance or physical qualification.
36/36 valid · 5 s grip 2/12, 4/12, 4/12 · grip + TCP stationarity 2/12, 0/12, 0/12 · simulation only
· Paired closed-loop evidence-gated probe and timer simulation study
A frozen 12-scene CPU study compares an authored native-evidence short-lift gate with a fixed-timer gate; both stop in a measured-position hold without full transport. All 24 attempts are valid, with 480,000 recorded 1 ms native rows and 24,000 commands; independent readback confirms the retained source and evidence unchanged. Evidence admitted 6/12 with zero false initial qualifications; the timer admitted 12/12 with eight false qualifications. Four evidence admissions and ten timer attempts were later invalidated, including two initially correct timer admissions. Two per variant remain in an admission hold. Final planning gates are 2/12 ready for simulated transport and 10/12 requiring another test lift. This finite closed-loop method comparison uses privileged simulator state and an arbitrary synthetic force ceiling; it is not an isolated classifier effect, learned policy, placement result or hardware qualification.
Four paired ABBA runs measured 15.70% lower mean wall compute and 12.60% lower mean process CPU compute after reducing allocations in the PiPER contact scan. A supplemental streaming audit exactly matched five baseline/candidate diagnostic pairs, including 200,010 recorded state snapshots, contacts and independently recomputed scorer state. The original reader-limit failure is retained. This is one synthetic scenario on a shared host; it establishes no new task learning, general performance gain, real-time response or hardware qualification.
Four ABBA timing runs · 5/5 dense pairs exactly matched · simulation only
· Native interface contracts and bounded CPU mechanics benchmark
Three pinned native simulation interfaces describe PiPER, SO-101, and UR5e with Robotiq 2F-85, preserving each arm and gripper command’s native units. A frozen 54-attempt roster covers three motion cases, three command periods, and two deterministic repeats; 21,654 state rows and 1,260 commands are retained, with zero audit mismatches. Twenty-one offline computation windows exceed their nominal command period; this is not a hardware timing measure. A data-only lift gate uses caller-declared bounds and is not connected to the controller. The existing PiPER policy remains 5/12 versus its authored teacher at 12/12; no new learning, cross-arm transfer, or physical qualification is claimed.
54/54 numerically valid · 21,654 state rows · 0 audit mismatches · simulation only
· Bounded multi-object OpenUSD task environment and verifier validation
A versioned recipe authors a static table, an open-top bin and one to four original rigid object instances around a pinned native PiPER. A frozen roster retains six accepted completion fixtures, six correctly rejected native counterexamples and four invalid layouts refused before stepping. An independent audit reconstructs all 13,262 physics observations and 1,337 replay frames with zero mismatches. The twelve native cases simulate 13.25 seconds in 16.550 wall seconds including construction and evidence outputs. PiPER holds home commands; rigid shoe mechanics remain assumptions. This establishes bounded environment and completion-verifier behavior, not learned manipulation, arbitrary USD import, camera accuracy or hardware transfer.
A native capture preview pairs genuine 160×120 RGB and metric-depth arrays with measured arm/finger state, causal authored decisions, issued references and final native controls. Two 20-second repeats of one fixed synthetic start match all 1,001 frame pairs and measured motion; a declared interruption retains two complete intervals with an unknown outcome. Independent decoding and byte readback cover all 2,005 registered camera observations. Four object-tracker channels remain missing, so strict training export refuses. Thirty-seven focused tests and a fresh installed-wheel native smoke pass. Depth precision, corrective demonstration quality, hard real-time performance, learned transfer and hardware qualification remain unproven.
Two identical 1,000-interval captures · two-step unknown prefix · tracker missing, training refused
· Bounded USD scene conversion and frozen passive physics checks
A real OpenUSD parser maps original rigid-object geometry, explicit mass, centred inertia, gravity and friction into a local PiPER/MuJoCo workstation. All 36 normal drop/sliding cases and nine negative controls pass a frozen CPU benchmark; 27/27 normal drops also satisfy an independent settling goal. Complete inputs, source, recorded native poses, failed-effect controls and explicit conversion limits remain inspectable. This is a bounded scene bridge, not arbitrary USD import, learned manipulation, calibrated shoe physics or hardware transfer.
45/45 physics checks · 36 normal cases and nine negative controls · simulation infrastructure
· Authored acquisition supervisor and recovered paired confirmation study
A finite 200 ms acquisition supervisor preserves frozen PiPER arm and jaw references while replacing indefinite learned waiting. A recovered study retains 508 scored confirmation attempts and four explicitly unknown interruptions across 512 scheduled requests; primary analysis conservatively imputes unknowns as failures. Strict completions are 42 bounded, 32 learned clock, 44 fixed and 41 teacher per 128 scheduled attempts. Both adjusted paired intervals include zero. All 556 complete development and confirmation attempts audit without mismatches; 22 explicit aborts prevent indefinite acquisition stalls, while retention and critical-contact failures remain. Compact summaries bind the locally retained raw archive; no new model, human-video skill or physical qualification is claimed.
A compact progress clock trained from 64 synthetic teacher episodes coordinates a frozen PiPER arm and jaw reference while acquiring a rigid block. All 512 confirmation attempts are valid: the learned clock completes 44/128 strict placements, its matched blind control 26/128, fixed timing 72/128 and the sampled teacher 67/128. The paired XY-group full-minus-blind interval spans −8.5 to +36.6 percentage points, while the learned clock trails fixed timing. An independent audit reconstructs all 768 main-study attempts with zero mismatches; separate pilot attempts remain public. Policy-call CPU p95 is 1.916 ms, with 728 full steps exceeding 20 ms on the shared host. No reliable feedback benefit, continuous learned recovery, human-video action learning or hardware qualification is established.
After five seven-command residual candidates each failed all 16 development cases, a separate 25-node learned jaw adapter preserves the frozen six-axis PiPER arm references. Trained on 64 synthetic teacher episodes and 64,000 causal labels, it completes 6/24 frozen confirmation cases, versus 8/24 for its matched blind control, 5/24 feedforward, 24/24 authored teacher and 0/24 open jaws. Feedback succeeds in 0/12 interrupted cases; 14/24 terminal-only passes conceal strict task failures. All 120 confirmation attempts are valid, and an independent numeric audit reconstructs 360 retained attempts. CPU inference p95 is 0.501 ms; this establishes no live-feedback advantage, recovered grip, human-video learning or hardware qualification.
A compact feedforward joint-and-jaw primitive trained on 32 successful synthetic demonstrations completes 5/12 frozen PiPER placement attempts from unseen starting positions. The authored teacher completes 12/12; three controls complete 0/12 each. All twelve learned terminal poses pass settling checks, while seven fail the full native contact-history gates. Complete tapes, failed development candidates, an independent audit and an installable local runner preserve the reliability gap. This is privileged-state reference distillation for a fixed rigid block and two trained goals, not human-video learning, learned feedback or physical robot qualification.
Learned 5/12 · teacher 12/12 · three controls 0/12 each · simulation only
· Paired synthetic benchmark for camera and drive evidence
A fresh eight-layout PiPER simulation benchmark adds separately processed RGB and drive checks to a local physical-goal verifier. The checks prevent 69 of 112 false completion readings while leaving 43 replay, identity, orientation, support and sampling blind spots; the conservative gate abstains on 73 of 80 baseline true completions, including all 64 deliberate sensor-fault controls. The measured CPU path is 22.1 ms median and 31.5 ms p95, excluding 200 ms acquisition. All raw channels and results are preserved; this is synthetic evidence, not hardware validation.
An offline audit traces the seven missed Jev completions without claiming an internal model cause. A local deterministic verifier then checks timestamped pose, motion, support, release and uncertainty on 200 fresh cases. It recognizes all 16 native positives and rejects all 48 native negatives; 128 evidence faults yield 120 abstentions and 8 incomplete answers. All 8 consistent forged streams produce physical false completions, exposing the remaining sensor-trust problem. CPU-only p95 is 0.318 ms, excluding the required 200 ms history and sensing. Complete native inputs, every decision, independent review and explicit failure cohorts remain public; no new Jev calls or hardware qualification.
The first measured run preserves all 64 requests from the image-goal v0.1 protocol. Jev returned valid structured responses for every trial and made no false-completion declaration, but recognized only 1/8 physically complete cases with privileged native telemetry and 0/8 with images alone. All four occluded complete scenes were called incomplete in both conditions. RGB abstention matched 8/16 ambiguous references. Median online latency was 1.776 seconds, p95 3.287 seconds. The release preserves every result, the independent review and the original unrun protocol. This small synthetic diagnostic does not qualify a robot completion gate, actuator loop or learned skill.
A frozen PiPER image-goal cohort crosses eight physical conditions with four layouts. Four moving/settled pairs have byte-identical images and different completion labels, exposing when one image cannot establish task completion. Native measurements verify 32 authored fixtures; a paired 64-request Jev protocol distinguishes RGB judgments from privileged simulator telemetry. The provider evaluation has not run because credentials were unavailable: accuracy, false-completion rates and service latency remain unmeasured. Every native state, image, frozen request hash and explicit unrun result is retained. No learned policy or hardware qualification is claimed.
32 native fixtures · 4 identical-image pairs · 64 Jev trials not run
· Research direction and passive native contact audit
A goal image specifies a destination; physical feedback verifies whether a robot gets there. We describe a bounded decision role for TypeSafe Jev, distinguish it from action-conditioned world-model planning, and propose an independently scored reaching experiment. A separate passive audit reproduces fourteen previously inspected rigid-box arm episodes exactly and records 249,502 native ticks. Millisecond contact gaps hidden between stored observations become visible, while task outcomes remain unchanged. Complete inputs, traces, source snapshots and a verified downloader preserve every case. This is measurement work, not a new policy benchmark, PiPER shoe result or hardware qualification.
A frozen 24-attempt PiPER diagnostic separates three rigid shapes, 30 g and 300 g loads, and mounted versus absent rake conditions. All attempts are valid; eight of twelve lightweight controls retain a direct lift while every 300 g attempt fails, including the small block. Removing the rake gains one narrow-shoe result and loses one block result, with no net completion gain. A verified settled reference, dense contact traces, unchanged drive fingerprint and independent scoring expose limits in the modeled grasp and force response. Recorded replays and runnable CPU code retain every failure. No autonomous tool removal, retrieval, learned skill or physical hardware qualification is claimed.
24/24 valid · 30 g: 8/12 lifts · 300 g: 0/12 · no net gain from removing rake
· Frozen native comparison with independent contact audit
A frozen twenty-attempt native MuJoCo comparison adds continuous acquisition checks and bounded lowering to the mounted-rake PiPER controller. Both policies complete the same 2/10 fresh cases: no gains or losses. Nine bilateral-contact-loss alarms follow a trace-verified native origin by 20–52 ms, median 30 ms; three acquisition failures are separate. Twelve supported lowering responses produce no additional completion. A non-finger-support gap remains unmonitored. Complete recorded trajectories, independent audit and development failures are public. Authored policies, synthetic contact and privileged observations establish no learned skill or physical robot qualification.
20/20 valid · 2/10 completed under both policies · 9 contact-loss alarms verified
· Two native environments with authored policies and recorded benchmark evidence
A frozen sixteen-attempt native MuJoCo study tests hinged shoe-box access and a mounted rake for an out-of-workspace shoe. Guarded access avoids two baseline lid collisions while task completion stays at 2/4; a transient withdrawal adds 0.70 simulated seconds without being necessary in its matched baseline. The mounted rake reaches an intermediate retrieval checkpoint in 3/4 cases, but all subsequent retained lifts fail and the short-tool case is deferred. All native traces and failures remain public. The browser replays recorded states; fresh physics runs locally. Privileged observations, synthetic contact and authored policies establish no learned skill or hardware qualification.
A six-axis PiPER and its coupled gripper respond to user targets while MuJoCo runs locally in a browser worker. The final native run retains 20 valid attempts: 14/16 target trials pass, both high-target repeats fail, and jaw, impulse, disabled-table and 600-second hold checks pass. A ten-minute headless Chromium run reports 51.389 render calls per second and a 0.9999795 simulation/wall-time ratio on the tested Apple M5 host; simulated movement response is 61.9 ms p95 across 61 commands. Render calls do not measure GPU presentation. Native/WASM numerical and loaded-contact-count checks pass, while strict raw contact counts differ in 13/1,500 states. All fifteen development shoe-lift attempts fail. Complete development, failure and timing records remain separate. This authored rigid-shoe environment establishes no learned skill, calibrated material model or physical robot qualification.
Six original MIT rigid shoe proxies compose with the pinned UR5e/Robotiq, PiPER and SO-101 models through versioned scene descriptors and offline native 3D previews. A frozen 54-attempt test records 27/27 supported pairs and 0/27 support successes when table collisions are disabled; all 54 gravity checks pass without numerical warnings. No loaded non-table shoe contacts appear in the 20 ms samples, which cannot rule out shorter transients. All sampled trajectories and 27,000 native-step durations remain available. On one Apple M5 host, 54 simulated seconds take 12.3616 wall seconds including fresh model construction; the slowest attempt takes 6.7047 seconds. Geometry and physical properties are authored assumptions. Loading, gravity and passive support do not establish shoe grasping, calibrated deformation, a learned policy or hardware transfer.
Six shoe proxies · three arms · support 27/27, disabled table 0/27 · manipulation unqualified
· Frozen paired reliability experiment with complete native evidence
Three frozen fresh PiPER cohorts retain 252 numerically valid attempts and 144 failed negative controls. A fixed diagonal grasp with bounded pose planning completes 23/36 placements versus 15/36 for frozen v0.4 and 18/36 with second-lift recovery alone: eight gained completions and none lost. The fifteen shared successes are 0.772 simulated seconds slower on average. Three lift-only second probes qualify and finish their tasks; the full grasp candidate needs none. Twelve low-friction acquisitions and one later retention failure remain. Every trajectory, native task outcome and raw computation duration is retained, including 1,035 windows exceeding 20 ms and a 3.082-second outlier on a loaded host. Native mechanics and scoring remain unchanged. Separate development keeps 200 earlier attempts, including discarded regressions. Authored privileged simulation establishes no learned skill, real-time guarantee or physical qualification.
A fresh frozen 84-attempt PiPER comparison completes 4/12 placements for v0.3 history, v0.4 carry-only and v0.4 full; all 48 matched open-jaw and hold controls fail. Two completed post-hold placements finish 2.440 and 2.640 simulated seconds sooner with unchanged 220/240 ms settling holds. Four low-friction failures reach refusal 2.88–2.90 seconds sooner without becoming successes. Fresh nominal-friction 60 g layouts expose two held-but-too-low resume failures in every mode; neither reaches speed restoration. Six earlier acquisition failures are traced to insufficient support under the pinned position servo and side pinch, not the actuator force cap. All native mechanics and outcome criteria remain unchanged. Complete trajectories and 84,000 raw computation windows are retained; none exceeds 20 ms on this host. Source-snapshotted development remains separate. Privileged authored simulation establishes no learned skill, true-slip accuracy or physical qualification.
All modes 4/12 · two resumed tasks faster · controls 0/48 · hardware unqualified
· Paired probe-only and contact-history experiment with complete confirmation receipt
A frozen paired PiPER experiment retains all 84 numerically valid attempts across twelve synthetic rigid-block scenes. History completes 6/12 placements versus 4/12 for both original motion and the identical acquisition probe without carry monitoring; all 48 separately matched open-jaw and hold controls fail. The two added completions occur at nominal friction with 60 g blocks. Four tasks complete after five closed holds/resumes, but only four of five resume windows meet the independent low absolute TCP-speed criterion. All six supported acquisition retries fail. Native gains, forces and outcome rules remain unchanged; requested movement reduction is distinct from actual speed, and two slowdown windows include early holds. Seventy-five of 84,000 computation windows exceed 20 ms. Complete sealed trajectories, original model notices, source-snapshotted development and pinned offline inspection preserve both success and failure. Privileged authored simulation establishes no learned skill, true-slip detector accuracy, shoe competence or physical qualification.
History 6/12 · baseline and probe-only 4/12 · controls 0/48 · hardware unqualified
· Paired authored controller experiment and complete confirmation receipt
A slow test lift and bounded supported retry add recorded acquisition decisions to the pinned PiPER contact pilot, but completion regresses: at moderate friction baseline completes 2/2 and guarded completes 0/2. Both complete 2/2 normal cases; neither completes the two low-friction cases. All 30 attempts remain numerically valid, including 18 failed controls and full-duration aborts. Four qualification decisions are followed by a brief independent following window, yet two tasks still fail; no successful recovery or overall improvement is demonstrated. Actual motion continues after hold requests, and 53/30,000 computation windows exceed 20 ms. Complete compressed receipts, original model notices and pinned offline inspection preserve the evidence. Authored privileged-state simulation establishes no detector accuracy, learned shoe skill, calibrated contact or physical robot qualification.
A pinned six-axis arm and coupled gripper complete 6/6 frozen synthetic rigid-block placements, while matched recorded-arm/open-jaw and hold controls complete 0/12. Native contact must produce grasp, sustained lift, retained carry, release and settled placement with tool retreat. All 18 attempts and timing traces remain in the receipt; one placement takes 29.51 wall seconds for 20 simulated seconds and misses 251 of 1,000 computation windows, so real-time operation is not established. An offline 3D mesh replay exposes the recorded motion. The authored privileged-state baseline provides a benchmark for a future compact DVIDIA action policy, not a learned footage-to-action model, validated shoe simulation or physical robot qualification.
A separate UR5e/Robotiq reference preserves pinned geometry, inertias, moving-link collision proxies and coupled gripper structure: fourteen modeled coordinates and seven commands. Eighteen deterministic unloaded motion/closure runs are numerically valid; at a 1 ms step, CPU physics throughput measures 40.80–47.29 simulated seconds per measured stepping wall second on one host, excluding loading, sensing, rendering, policy and real actuation. Adjacent timestep comparisons and a matched static overlap expose numerical sensitivity and collision participation. A pure-Python paired trace scorer measures joint, tool and jaw mismatch without automatic qualification. Left/right five-finger references are separately audited; LeRobot collection with an original DVIDIA policy is proposed. Serial calibration, physical response, shoe task learning and bimanual execution remain unmeasured. An offline 3D mesh replay now exposes the recorded arm/gripper poses without rerunning physics. PiPER and SO-101 now have pinned CPU profiles with 6/6 valid unloaded numerical attempts each and explicit native command units. PiPER also has a frozen authored rigid-block contact pilot: nominal placement 2/2, low-friction placement 0/1, open-jaw 0/3 and hold 0/3, with all nine attempts numerically valid. One of 9,000 computation windows exceeds 20 ms; hardware real-time response is unqualified. SO-101 object tasks, learned policies and physical qualification remain open. YAM, OpenArm, ARX X5 and PAROL6 remain upcoming.
Pinned CPU mechanics · PiPER contact pilot · physical fidelity unmeasured
· Local observation workflow and engineering feasibility pilot
DVIDIA Observation Lab supplies local object/palm proposals, reviewed timeline revisions and separate causal image-motion forecasts; real annotation and motion accuracy remain unmeasured. A new CPU pilot learns next-20-ms joint/jaw deltas from current native state and issued commands. Twelve seeded groups retain 36 recordings with 8/2/2 train/development/test grouping; 4,553 test transitions are not independent robot trials. The 6,438-byte action model computes at 0.207 ms p95 on one host and improves RMSE, while jaw MAE/worst error regress against persistence and an initial opening is substantially overpredicted. These are privileged simulator measurements and model-only timings. A proposed shoe capture plan has no footage or executed shoe benchmark. The v0.1 release and hosted synthetic review demo remain separate; no shoe policy, physical adapter, validated shoe physics or hardware qualification is established.
Local review release · motion and actuator source pilots · hardware unqualified
· Architecture proposal and staged validation roadmap
An updated architecture connects immutable Skillspace recordings, reviewed task-label proposals and synchronized measured telemetry to grouped datasets, separate visual and motor learners, and a proposed local state supervisor. The implemented CPU baseline retains authored phases, gripper decisions and recovery around its learned joint-delta head. New annotation and supervisor pilots, fresh frozen confirmation, physics calibration and compatible skill-installation gates define the next milestones. Automatic labeling, learned supervision, human-video action recovery and physical arm qualification remain unestablished; this publication adds no new training or hardware result.
Architecture proposal · current simulation alpha · hardware unqualified
· Product implementation and measured synthetic pilot
The public Dvidia-Inference/dvidia-training repository provides a standalone CPU pipeline that audits Skillspace media, freezes grouped splits and exports inspectable artifacts. Ten synthetic six-second clips train a visual predictor with a 0.5106-second median pipeline time across three repeats on one macOS host. Aligned native simulator telemetry supplies 4,029 training labels for a learned joint-delta head, evaluated on one held-out recording. Its capsule passes native placement qualification in that recording's exact scene; identical joint targets with open jaws fail. Installation and hardware guides accompany the package. Human-video action recovery, shoe organization, demonstration sufficiency, independent scene generalization and physical robot competence remain unestablished.
Public CPU package · one native scene qualified · simulation-only
· Learned movement artifact and installation prototype
A compact learned CPU movement head replaces joint-target computation inside an authored placement supervisor. Teacher and student both pass 7/8 in a new frozen evaluation, while episode wall time increases from 6.273 to 6.360 seconds. A 39,845-byte JSON capsule imports into the local lab, checks its exact runtime and qualifies the current scene through native placement and a recorded-arm/open-jaw control. Scene changes require a fresh test. One known scene and repeat pass native network denial; perception, complete task learning, other embodiments and hardware remain unqualified.
Simulation-only · teacher/student 7/8 · no overall speedup
· Paired simulation benchmark and product assessment
Two controller revisions improve development completion from 14/20 to 17/20, but the frozen paired test declines from 30/32 to 29/32: both pass all 24 normal-strength cases while actuator-stress completion falls from 6/8 to 5/8. Instrumented CPU throughput also declines. We publish the regression, unchanged contact and success rules, exact controller snapshots, richer physical inputs, runtime-bound installation and a proposed staged Skillspace contract. Contact forces and penetration remain uncalibrated; learned perception and physical qualification are open.
Mixed held-out result · 30/32 baseline, 29/32 updated
· Product prototype and native contact qualification
The first link-to-arm product path installs an existing DVIDIA placement envelope, binds a local authored controller to one simulated arm, and reuses it in independently grounded scenes. Native jaw contacts carry a free box; six layouts complete, while no-motion and identical joint commands with open jaws each complete zero of six. Installed execution passes process-level offline checks. Task metadata contains no learned policy; perception, cup geometry, full arm collisions and physical qualification remain open.
Public skill link to authored adapter · 6/6 scene placements
The first runnable product loop combines an anchored cable environment, local controller-parameter search, frozen evaluation, offline replay and a manifest-backed skill pack. The selected controller and a fixed feedback baseline both completed 10/10 held-out simulated episodes; zero-force and release controls completed 0/10. A native macOS network-denial run completed 3/3 additional episodes. Ideal endpoint attachment and privileged state bound the result; physical rope calibration, gripping, sensors and GPU throughput remain untested.
CPU simulation-only · Controller search and frozen evaluation executed
A design for an offline, self-hosted training environment that turns demonstrations into calibrated practice and qualified robot skills. We examine contact, deformation, sensor feedback and the performance–accuracy tradeoff, then report 54 deterministic native MuJoCo CPU fixture trials. Fine endpoint agreement did not establish accurate contact transients, and coarse timesteps missed collisions. Minimum GPU memory, rope accuracy and physical transfer remain unmeasured.
54 CPU fixture trials · GPU and physical transfer untested
Three Grok 4.7 research passes surface observation-to-action, synthetic-practice and reusable-repair ideas. Independent primary-source checks compare those findings with our thesis: observation supplies goals and state, language retrieves a skill, spatial expectations guide attention, and executable feedback with evaluation establishes competence. X posts remain discovery leads. No universal demonstration count or arbitrary-robot transfer follows from the reviewed results.
A synthesis of learning from human demonstrations, language-conditioned spatial attention, simulation expansion and robot-specific skill delivery. Spatial priors can accelerate inspection but do not determine contact state or control. The proposed downloadable skill packages a controller, initiation and termination conditions, compatible sensing and hardware, recovery behavior and measured evidence. Separate experiments would test attention, data coverage, feedback and transfer.
A preregistered image-API feasibility run processed five licensed three-second POV excerpts using four timestamped frames, a Gemma caption and a Clef-flash check. All five completed at a price-derived inference estimate of $0.00140634. Post-run inspection found an object error accepted by the checker and omitted state changes. API and schema compatibility were demonstrated; annotation accuracy and whole-task success were not.
Five licensed excerpts · Accuracy evaluation outstanding
An Argus review and a proposal to connect original recordings, reviewed episodes, immutable datasets, training runs, held-out evaluation and robot-tested releases. The review separates video annotation from action supervision and defines compatibility beyond a robot’s degree-of-freedom count. A proposed 40-episode annotation pilot and evidence-based graduation criteria remain unrun.
A research agenda covering visible capture evidence, relations over time, evaluation provenance and physical grounding. Each question has a bounded protocol and a failure condition. The aim is to measure whether structure and physical signals improve a defined task before scaling collection. Strategic opinions, author-reported results and our untested proposals are kept distinct.
Four baseline protocols proposed
No publications match this search.
Research footprints
Construct scenes and test the whole finish
An editable local recipe places up to four rigid objects around PiPER and an open bin. Six completion fixtures pass, six native counterexamples stay incomplete and four invalid layouts are refused. An independent audit checks 13,262 physics observations and 1,337 replay frames with zero mismatches. The arm holds home; this is environment infrastructure, not learned placement or hardware qualification.
Connect camera frames to the commands the arm received
Native preview 0.1.3.dev1 captures actual RGB-D, measured joints and signed finger velocities beside issued, ramped and applied references. Two full 1,000-interval runs repeat one fixed start with identical frames and motion; interruption preserves a two-step unknown prefix. All 37 focused tests and an isolated installed-wheel capture pass. Missing object-tracker channels still refuse training export, and depth precision remains unverified. This is synchronized evidence collection, not a new learned skill.
Preserve the motor evidence before fitting the next learner
SDK 0.1.2 adds a causal PiPER data contract and installed Terminal validator/exporter. All 203 required default-branch tests pass, including 27 recorder tests. Native channel order, gripper units, missingness, interruptions and outcome separation remain explicit. The example is synthetic metadata; live capture, corrective demonstrations and policy training remain separate work.
Bound acquisition waiting and preserve interrupted outcomes
A 200 ms authored supervisor completes 42/128 scheduled placements versus 32/128 learned clock, 44/128 fixed and 41/128 teacher. Both paired comparisons are inconclusive. Recovery preserves the original 72 receipts and four unknowns, running only 436 never-started requests. All 556 complete development and confirmation attempts audit without mismatches; retention and critical contacts remain unresolved.
A bounded USD recipe maps rigid geometry, mass, centred inertia, gravity and friction into a pinned PiPER workstation. All 45 passive checks pass, including nine disabled-effect controls; 27/27 normal drops also satisfy independent settling goals. Every input, recorded native pose and read-only audit is retained. Arbitrary asset import, learned manipulation and physical calibration remain open.
A learned timing model completes 44/128 strict placements versus 26/128 for its matched blind clock, 72/128 for fixed progression and 67/128 for the acquisition teacher. Its advantage over blind is inconclusive and it trails fixed timing. All 768 main-study attempts audit without mismatches. The complete evidence and recorded comparison retain failed attempts; action corrections, camera-based control and physical transfer remain open.
Learn a first PiPER placement primitive from demonstrations
A compact feedforward model trained on 32 synthetic teacher demonstrations completes 5/12 frozen placement attempts from unseen starts. The teacher completes 12/12; three controls complete none. All twelve learned final poses settle near their targets, but seven fail full contact-history gates. The recorded viewer, complete evidence and installable local runner preserve the reliability gap. Learned feedback, human-video learning and physical transfer remain open.
All 64 frozen requests returned valid responses. No false completion was observed, but images plus telemetry recognized only 1/8 completed states, and images alone recognized 0/8. Occluded finished scenes failed in both conditions. Median hosted latency was 1.776 seconds. The original protocol remains unchanged; all measured cases and independent review are public.
Thirty-two native PiPER fixtures include four moving/settled pairs with identical images and different physical completion labels. The frozen 64-request Jev protocol compares RGB with privileged simulator telemetry. Jev has not been run; all provider trials remain explicitly unrun, with accuracy and service latency unmeasured. Native states, images, source hashes and physical labels are public.
Connect visual goals to measured physical feedback
Fourteen passive rigid-box arm replays reproduce every original sampled state and outcome, recording 249,502 native ticks. Millisecond contact gaps appear between stored observations; a successful comparator also has gaps. The separate architecture note evaluates Jev as a candidate decision layer and defines a goal-image reaching test. No visual model is integrated and no controller is promoted.
Separate object difficulty from the gripper and attached tool
A frozen direct-grasp diagnostic retains 24 valid attempts. Lightweight controls complete 8/12 lifts; every 300 g condition fails, including the small block. Removing the rake gains one narrow-shoe result and loses one block result. Relative lift, settled references and dense contact traces make the modeled drive and geometry limits visible. These are diagnostic controls, not a shoe retrieval rate or hardware qualification.
Measure contact-loss response without claiming a better grasp
Twenty valid fresh attempts complete the same 2/10 cases under both controllers. Nine trace-verified bilateral-contact-loss alarms arrive in 20–52 ms; twelve supported lowering responses produce no additional completion. Full trajectories and an independent audit expose failed acquisition and an unmonitored support-contamination gap. Hardware transfer remains open.
Make skill prerequisites and failed recovery measurable
Two native environments retain sixteen valid held-out attempts. A lid guard avoids two baseline contacts while completion remains 2/4. A mounted rake reaches three of four intermediate retrieval checkpoints, but no subsequent retained lift succeeds. Recorded replays expose every failure; fresh physics runs locally. Tool acquisition, learned control and hardware transfer remain open.
Drag a target and nudge a free rigid trainer with live browser MuJoCo physics. The final native run keeps 20 valid attempts: 14/16 targets pass, both high-target repeats fail, and four mechanics checks pass. A ten-minute browser run reports 51.39 render calls per second on the tested host. Strict raw-contact parity fails in 13/1,500 sampled states despite numerical agreement; all fifteen development shoe-lift attempts fail. Learned shoe skills, calibrated contact and hardware transfer remain open.
Six original rigid shoe proxies now compose with UR5e/Robotiq, PiPER and SO-101. Versioned scene files, offline 3D previews and a frozen 54-attempt benchmark retain every trajectory: supported pairs 27/27, disabled-table support 0/27, all gravity checks pass. Native loading and passive contact are tested; shoe grasping, deformation and physical transfer remain unqualified.
Complete more placements with a bounded second lift and revised grasp
Three fresh cohorts retain 252 valid attempts. The revised grasp completes 23/36 placements versus 15/36 for v0.4 and 18/36 with second lift alone; all 144 controls fail. Eight completions are gained and none lost, while shared successes average 0.772 simulated seconds slower. Twelve low-friction acquisitions and one retention failure remain. All 1,035 computation overruns are retained; physical qualification is open.
Separate faster recovery from an ineffective retry
A fresh 84-attempt comparison completes 4/12 placements in every mode; all 48 controls fail. Two resumed tasks finish 2.440–2.640 simulated seconds sooner after unchanged settling holds, while four failed grips are refused 2.88–2.90 seconds sooner. Eight failures remain, including two held blocks below the resume-height gate. Every native trajectory and raw computation duration is retained; physical qualification is open.
A frozen 84-attempt paired cohort completes 6/12 history placements versus 4/12 original and probe-only placements; all 48 controls fail. Four tasks complete after five closed holds/resumes, but all six supported acquisition retries fail. Only four of five resume windows meet the independent low absolute tool-speed check. Actual motion and 75 missed computation windows remain visible; physical robot qualification is open.
A slow probe and one supported retry retain 30 valid attempts. At moderate friction baseline completes 2/2 while guarded completes 0/2; normal cases both pass and low-friction cases both fail. No successful recovery or overall improvement is demonstrated. Four brief following windows do not establish detector accuracy; 53 computation windows miss 20 ms. Every full attempt remains public.
A frozen authored rigid-block pilot retains nine valid attempts: nominal placement 2/2, low-friction placement 0/1, open-jaw 0/3 and hold 0/3. One of 9,000 computation windows exceeds 20 ms. Complete task receipts remain distinct from unloaded profiles; learned skills, shoe physics and physical qualification are open.
PiPER and SO-101 join the library with pinned models, explicit arm/gripper units, preserved licenses and complete numerical receipts. Each completes six unloaded motion checks offline. Their learned-policy and physical qualification remain open. The PiPER contact pilot and UR5e grasp demonstration have separate task receipts.
A free rigid block must be grasped, lifted, carried, released and left settled at its destination. Frozen scenarios compare the authored controller with identical arm commands and open jaws, and with a hold control. Full traces and 3D replay preserve the evidence for the next compact action-learning experiment. Shoe physics, a learned task policy and hardware transfer remain open.
A pinned six-axis arm and coupled gripper expose fourteen coordinates and seven commands. All 18 unloaded numerical runs are valid; 1 ms CPU physics measures 40.80–47.29× elapsed-time throughput with sensing, rendering and policy excluded. A paired trace scorer makes physical mismatch measurable. Five-finger references and LeRobot collection have separate contracts; hardware fidelity and shoe competence remain unmeasured.
A 6,438-byte CPU model predicts the next 20 ms of joint/jaw movement from native state and issued commands. Twelve seeded groups retain 36 recordings; 4,553 test transitions are not independent trials. Warm computation measures 0.207 ms p95, but jaw mean absolute/worst errors regress against persistence and the first opening is overpredicted. Shoe collection, hardware sensing and validated shoe physics remain open.
A manually seeded CPU source alpha records pixel speed and causal forecasts. Six authored cases process 720 frames in 1.454 seconds; occlusion and low-texture losses remain visible, and a 500 ms reversal forecast misses by nearly 90 pixels. The v0.1 release/demo remain separate; real-scene accuracy, metric speed and live robot control are unqualified.
A diagram connects footage review, measured robot state, local training, frozen simulation evaluation and compatible skill installation. Automatic annotation and a learned task supervisor are proposed; their companion validation studies and physical robot qualification remain unrun.
One-click local installation binds a DVIDIA placement envelope to an authored contact controller. Six scene layouts pass; identical joint commands with open jaws and no-motion controls each pass 0/6. Physical transfer and learned policies remain open.
A local controller-training loop and simulation-only skill pack now run on CPU. Trained and baseline feedback each pass 10/10 simulated seeds; zero force and early release each pass 0/10. Real gripping, calibration and GPU qualification remain open.
54 deterministic CPU fixture trials recorded. Endpoint agreement coexists with unqualified force transients; coarse wall tests expose missed contact. No GPU or physical-transfer claim.
Compare the observation-to-skill thesis with prior work
Spatial attention, demonstration coverage, executable simulation and reusable repairs are reviewed separately. Three parallel Grok 4.7 searches support discovery; primary sources support the claims.
Four baseline protocols and a Skillspace graduation proposal establish the questions to test. A proposed protocol is distinct from an executed experiment.