---
schema_version: "1.0.0"
edition_id: "mp-2026-10-22-morning-0104"
published_at: "2026-10-22T09:00:00.000-04:00"
modified_at: "2026-10-22T09:00:00.000-04:00"
canonical_url: "https://themachinepress.com/edition/2026-10-22"
story_count: 27
lead_story_id: "mp-2026-10-22-001"
---

# The Machine Press — Morning edition

Edition ID: `mp-2026-10-22-morning-0104`  
Published: 2026-10-22T09:00:00.000-04:00  
Canonical edition: https://themachinepress.com/edition/2026-10-22

Four 80-centimeter superconducting links carried state transfer, remote two-qubit gates and simultaneous Bell pairs between separate processor modules.

## 1. Four Cables Bridge Two 120-Qubit Processors {#mp-2026-10-22-001}

- Story ID: `mp-2026-10-22-001`
- Type: `lead`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-001/four-cables-bridge-two-120-qubit-processors

**Dek:** Four 80-centimeter superconducting links carried state transfer, remote two-qubit gates and simultaneous Bell pairs between separate processor modules.

Modular quantum computers need links that preserve coherent operations across more than one chip. Researchers demonstrated four long-distance couplers between two 120-qubit superconducting processors. Each link used two tunable coupler qubits, RF-SQUID interfaces and 80 centimeters of aluminum coax. Relay-based state transfer averaged 99.00 percent network-benchmarking fidelity, while distant controlled-Z gates averaged 94.75 percent and reached 96.42 percent on the best link. The team also generated entangled states across the modules, including simultaneous Bell pairs on adjacent couplers. These are component-level results under the reported setup, not a fault-tolerant system, but they show a concrete route for scaling modular machines beyond a single processor package.

### Why it matters {#why-it-matters-mp-2026-10-22-001}

Four 80-centimeter superconducting links carried state transfer, remote two-qubit gates and simultaneous Bell pairs between separate processor modules.

### Limits and context {#limitations-mp-2026-10-22-001}

- These are component-level results under the reported setup, not a fault-tolerant system, but they show a concrete route for scaling modular machines beyond a single processor package.

### Claims and sources {#claims-mp-2026-10-22-001}

- Four 80-centimeter superconducting links carried state transfer, remote two-qubit gates and simultaneous Bell pairs between separate processor modules. [source-2026-10-22-001] — Qualification: These are component-level results under the reported setup, not a fault-tolerant system, but they show a concrete route for scaling modular machines beyond a single processor package.

## 2. The Summary Forgot Who Said What {#mp-2026-10-22-002}

- Story ID: `mp-2026-10-22-002`
- Type: `secondary`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-002/the-summary-forgot-who-said-what

**Dek:** Five coding-agent harnesses flattened message roles before compaction, letting forged approvals survive as memory; four role-typed harnesses resisted the same test.

Long-running coding agents summarize old context to keep working. The study identifies a trust boundary before that summary is written: several harnesses first serialize typed messages into one plain string with textual role labels. Untrusted content can then imitate a user turn, and the summarizer no longer has authoritative provenance to distinguish it. Across five open-source flattening harnesses, forged approvals survived compaction 87 to 100 percent of the time; the same attack succeeded zero percent across four harnesses that retained typed roles. In follow-on tests, laundered instructions drove sandboxed command execution and could steer reviews away from a planted backdoor. A code-level provenance check caught 77 percent of laundering while demoting 2 percent of legitimate summary content, and escaping role labels offers a simpler serializer fix. The results concern the tested harness designs, not every coding agent.

### Why it matters {#why-it-matters-mp-2026-10-22-002}

Five coding-agent harnesses flattened message roles before compaction, letting forged approvals survive as memory; four role-typed harnesses resisted the same test.

### Limits and context {#limitations-mp-2026-10-22-002}

- The results concern the tested harness designs, not every coding agent.

### Claims and sources {#claims-mp-2026-10-22-002}

- Five coding-agent harnesses flattened message roles before compaction, letting forged approvals survive as memory; four role-typed harnesses resisted the same test. [source-2026-10-22-002] — Qualification: The results concern the tested harness designs, not every coding agent.

## 3. The World Model Starts Writing Reusable Skills {#mp-2026-10-22-003}

- Story ID: `mp-2026-10-22-003`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-003/the-world-model-starts-writing-reusable-skills

**Dek:** AW-0 couples code, future video and robot actions so predicted behavior can become a reusable program.

AW-0 models code, video and actions as three linked generative streams. A code expert produces parameterized routines and control logic while video and action experts predict future states and motor commands. An automated pipeline labels human and robot trajectories with programs and consolidates repeated routines into a skill library. Evaluations on RoboTwin 2.0 and physical bimanual tasks report gains over compared vision-language-action and world-action baselines. Those results are specific to the tested tasks, but they connect visual prediction with compositional execution.

### Why it matters {#why-it-matters-mp-2026-10-22-003}

AW-0 couples code, future video and robot actions so predicted behavior can become a reusable program.

### Limits and context {#limitations-mp-2026-10-22-003}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-10-22-003}

- AW-0 couples code, future video and robot actions so predicted behavior can become a reusable program. [source-2026-10-22-003]

## 4. Interactive World Models Need Their Own Compression Trade-off {#mp-2026-10-22-004}

- Story ID: `mp-2026-10-22-004`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-004/interactive-world-models-need-their-own-compression-trade-off

**Dek:** CADRE tests causal video autoencoders across spatial and temporal compression instead of inheriting assumptions from offline video generation.

Interactive world models pay for every latent frame in response time. CADRE holds most components constant while varying spatial and temporal compression across manipulation and navigation settings. The study finds different optima: temporal compression costs less when a manipulation camera stays still, while spatial compression is more useful when navigation moves the camera. At high compression, a diffusion decoder was sharper than perceptual regression. The result is a latency-quality study across four environments, not a universal recipe for every simulator.

### Why it matters {#why-it-matters-mp-2026-10-22-004}

CADRE tests causal video autoencoders across spatial and temporal compression instead of inheriting assumptions from offline video generation.

### Limits and context {#limitations-mp-2026-10-22-004}

- The result is a latency-quality study across four environments, not a universal recipe for every simulator.

### Claims and sources {#claims-mp-2026-10-22-004}

- CADRE tests causal video autoencoders across spatial and temporal compression instead of inheriting assumptions from offline video generation. [source-2026-10-22-004] — Qualification: The result is a latency-quality study across four environments, not a universal recipe for every simulator.

## 5. One Exploratory Clip Calibrates the World Model {#mp-2026-10-22-005}

- Story ID: `mp-2026-10-22-005`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-005/one-exploratory-clip-calibrates-the-world-model

**Dek:** PhysiCalWorld compresses a test-scene interaction into a reusable descriptor for mass, damping and friction effects.

A short visual history may not reveal the physical properties that govern what happens next. PhysiCalWorld encodes one exploratory interaction from the test scene into a compact descriptor token, then conditions later action-driven rollouts on that token without fine-tuning or explicit physics labels. Across four environments with varied dynamics, the authors report better prediction and visual-planning success, including outside the training ranges. The learned token is an empirical latent description, not a measured physical parameter set.

### Why it matters {#why-it-matters-mp-2026-10-22-005}

PhysiCalWorld compresses a test-scene interaction into a reusable descriptor for mass, damping and friction effects.

### Limits and context {#limitations-mp-2026-10-22-005}

- A short visual history may not reveal the physical properties that govern what happens next.
- The learned token is an empirical latent description, not a measured physical parameter set.

### Claims and sources {#claims-mp-2026-10-22-005}

- PhysiCalWorld compresses a test-scene interaction into a reusable descriptor for mass, damping and friction effects. [source-2026-10-22-005] — Qualification: A short visual history may not reveal the physical properties that govern what happens next.

## 6. The Robot Predicts Force Alongside Motion {#mp-2026-10-22-006}

- Story ID: `mp-2026-10-22-006`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-006/the-robot-predicts-force-alongside-motion

**Dek:** CW-WAM jointly forecasts video, movement and contact wrench, then feeds force references to a faster controller.

Vision alone cannot say how hard a robot is pressing. CW-WAM predicts aligned motion and wrench chunks and keeps a fixed history of earlier forces. A 20-hertz admittance controller follows those references between slower model queries. Compared with a version without wrench input, reported insertion success rose from 25 to 70 percent for square pegs and from 40 to 95 percent for power adapters, while force-reference error also fell on tested tasks. The gains are benchmark and hardware results under the authors' setup.

### Why it matters {#why-it-matters-mp-2026-10-22-006}

CW-WAM jointly forecasts video, movement and contact wrench, then feeds force references to a faster controller.

### Limits and context {#limitations-mp-2026-10-22-006}

- Vision alone cannot say how hard a robot is pressing.

### Claims and sources {#claims-mp-2026-10-22-006}

- CW-WAM jointly forecasts video, movement and contact wrench, then feeds force references to a faster controller. [source-2026-10-22-006] — Qualification: Vision alone cannot say how hard a robot is pressing.

## 7. The Student Learns to Search While Uncertain {#mp-2026-10-22-007}

- Story ID: `mp-2026-10-22-007`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-007/the-student-learns-to-search-while-uncertain

**Dek:** On-policy student histories let a privileged teacher train exploration without the teacher needing to explore.

ASTEROID lets a partially observed student collect its own history, then asks a privileged expert for actions conditioned on that history. The student therefore learns when to gather information and when to exploit what it found. Across eleven environments, the paper reports roughly double the returns under fixed exploration budgets and about one hundredth the interactions of compared reinforcement-learning baselines. In a vision-denied robot transfer, success rose 40 percent over standard distillation. These are controlled evaluations, not proof of general exploration across open worlds.

### Why it matters {#why-it-matters-mp-2026-10-22-007}

On-policy student histories let a privileged teacher train exploration without the teacher needing to explore.

### Limits and context {#limitations-mp-2026-10-22-007}

- These are controlled evaluations, not proof of general exploration across open worlds.

### Claims and sources {#claims-mp-2026-10-22-007}

- On-policy student histories let a privileged teacher train exploration without the teacher needing to explore. [source-2026-10-22-007] — Qualification: These are controlled evaluations, not proof of general exploration across open worlds.

## 8. Biomedical Vision Gets a Batch-Shift Scoreboard {#mp-2026-10-22-008}

- Story ID: `mp-2026-10-22-008`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-008/biomedical-vision-gets-a-batch-shift-scoreboard

**Dek:** BatchBench tests whether image models survive unseen experimental batches in the large JUMP Cell Painting collection.

Biomedical images can change with laboratory, plate, instrument and acquisition conditions even when the biological question stays the same. BatchBench turns those shifts into a public robustness leaderboard built on JUMP Cell Painting. The benchmark separates familiar from unseen experimental batches and data sources so gains cannot be mistaken for ordinary in-distribution accuracy. It supplies evaluation infrastructure rather than a clinical model, and performance on Cell Painting does not establish diagnostic reliability.

### Why it matters {#why-it-matters-mp-2026-10-22-008}

BatchBench tests whether image models survive unseen experimental batches in the large JUMP Cell Painting collection.

### Limits and context {#limitations-mp-2026-10-22-008}

- Biomedical images can change with laboratory, plate, instrument and acquisition conditions even when the biological question stays the same.
- The benchmark separates familiar from unseen experimental batches and data sources so gains cannot be mistaken for ordinary in-distribution accuracy.
- It supplies evaluation infrastructure rather than a clinical model, and performance on Cell Painting does not establish diagnostic reliability.

### Claims and sources {#claims-mp-2026-10-22-008}

- BatchBench tests whether image models survive unseen experimental batches in the large JUMP Cell Painting collection. [source-2026-10-22-008] — Qualification: Biomedical images can change with laboratory, plate, instrument and acquisition conditions even when the biological question stays the same.

## 9. Long Robot Tasks Get Atomic Credit {#mp-2026-10-22-009}

- Story ID: `mp-2026-10-22-009`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-009/long-robot-tasks-get-atomic-credit

**Dek:** AtomicRL prevents failures in one subtask from contaminating credit across the whole manipulation sequence.

Long-horizon robot policies often receive one delayed reward, obscuring which subtask failed. AtomicRL gates temporal-difference residuals at semantic task boundaries and targets exploration at the least-mastered atomic step. On LIBERO-LONG, the paper reports 3.6- and 4.0-point gains over PPO across two VLA architectures; on CALVIN ABC-D, full-task performance improved by as much as 9.5 points. A Real2Sim2Real pipeline raised average real-world task performance by 21.3 percent in the reported tests.

### Why it matters {#why-it-matters-mp-2026-10-22-009}

AtomicRL prevents failures in one subtask from contaminating credit across the whole manipulation sequence.

### Limits and context {#limitations-mp-2026-10-22-009}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-10-22-009}

- AtomicRL prevents failures in one subtask from contaminating credit across the whole manipulation sequence. [source-2026-10-22-009]

## 10. An Earth-Trained Humanoid Stumbles on the Moon {#mp-2026-10-22-010}

- Story ID: `mp-2026-10-22-010`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-010/an-earth-trained-humanoid-stumbles-on-the-moon

**Dek:** Simulations expose asymmetric transfer across gravity, terrain and soil—and show one randomized policy that remained broadly stable.

On a simulated Unitree G1, Earth-trained policies retained 0.64 to 0.75 survival on flat rigid lunar ground, while Moon-trained policies could not stand when moved back to Earth. A released generalist controller reached at most 0.10 survival on the Moon. Rough-terrain training helped, and one policy randomized across gravity on resistive-force terrain achieved at least 0.84 mean survival at Earth gravity and 0.96 at lunar gravity across key tests. Every lunar result depends on simulation, contact assumptions and modeled soil stiffness.

### Why it matters {#why-it-matters-mp-2026-10-22-010}

Simulations expose asymmetric transfer across gravity, terrain and soil—and show one randomized policy that remained broadly stable.

### Limits and context {#limitations-mp-2026-10-22-010}

- On a simulated Unitree G1, Earth-trained policies retained 0.64 to 0.75 survival on flat rigid lunar ground, while Moon-trained policies could not stand when moved back to Earth.

### Claims and sources {#claims-mp-2026-10-22-010}

- Simulations expose asymmetric transfer across gravity, terrain and soil—and show one randomized policy that remained broadly stable. [source-2026-10-22-010] — Qualification: On a simulated Unitree G1, Earth-trained policies retained 0.64 to 0.75 survival on flat rigid lunar ground, while Moon-trained policies could not stand when moved back to Earth.

## 11. A Fallen Humanoid Chooses How to Rise {#mp-2026-10-22-011}

- Story ID: `mp-2026-10-22-011`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-011/a-fallen-humanoid-chooses-how-to-rise

**Dek:** RISE steers heading, speed and path at inference time by planning future body states before executing a recovery.

The same fallen pose can support several safe ways back to standing. RISE learns a vocabulary that links intended movements to low-level actions, then uses a visual diffusion planner to predict future body states. Task costs steer that prediction toward a requested heading, rise speed or route without retraining. The authors demonstrate varied terrain recoveries in simulation and on a Unitree G1. The work controls among learned recovery behaviors; it does not guarantee recovery from arbitrary falls.

### Why it matters {#why-it-matters-mp-2026-10-22-011}

RISE steers heading, speed and path at inference time by planning future body states before executing a recovery.

### Limits and context {#limitations-mp-2026-10-22-011}

- The work controls among learned recovery behaviors; it does not guarantee recovery from arbitrary falls.

### Claims and sources {#claims-mp-2026-10-22-011}

- RISE steers heading, speed and path at inference time by planning future body states before executing a recovery. [source-2026-10-22-011] — Qualification: The work controls among learned recovery behaviors; it does not guarantee recovery from arbitrary falls.

## 12. The Implant Looks Plausible but Cannot Be Placed {#mp-2026-10-22-012}

- Story ID: `mp-2026-10-22-012`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-012/the-implant-looks-plausible-but-cannot-be-placed

**Dek:** A zero-shot audit found no implantable output from three released shape-completion models across 239 cranial defects.

Researchers tested three pretrained shape-completion systems on 210 synthetic cranial defects and 29 real post-craniectomy cases. None of the zero-shot outputs both avoided intersecting remaining bone and added useful material; a task-specific volumetric model was implantable on all but one case. The authors also warn that point outputs do not encode implant thickness. The audit does not rule out future models or specialist workflows, but it shows that visually plausible completion is not enough for clinical use.

### Why it matters {#why-it-matters-mp-2026-10-22-012}

A zero-shot audit found no implantable output from three released shape-completion models across 239 cranial defects.

### Limits and context {#limitations-mp-2026-10-22-012}

- The authors also warn that point outputs do not encode implant thickness.
- The audit does not rule out future models or specialist workflows, but it shows that visually plausible completion is not enough for clinical use.

### Claims and sources {#claims-mp-2026-10-22-012}

- A zero-shot audit found no implantable output from three released shape-completion models across 239 cranial defects. [source-2026-10-22-012] — Qualification: The authors also warn that point outputs do not encode implant thickness.

## 13. Document Ingestion Erased the Author's Emphasis {#mp-2026-10-22-013}

- Story ID: `mp-2026-10-22-013`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-013/document-ingestion-erased-the-author-s-emphasis

**Dek:** Mainstream converters dropped color and highlighting cues, turning singled-out policy requirements into confident misses.

Using 76 real policy documents and deterministic labels from their OOXML properties, the study asks agents to recover requirements authors visually singled out. Formatting-blind ingestion fell from 0.690 micro-F1 to 0.150 with one converter and 0.075 as plain text. Highlight and color accounted for nearly all of the measured loss, and color was preserved by none of the mainstream converters tested. The pattern replicated in PDF and across six answering models. The result concerns visual emphasis in this corpus, not every document or requirement.

### Why it matters {#why-it-matters-mp-2026-10-22-013}

Mainstream converters dropped color and highlighting cues, turning singled-out policy requirements into confident misses.

### Limits and context {#limitations-mp-2026-10-22-013}

- The result concerns visual emphasis in this corpus, not every document or requirement.

### Claims and sources {#claims-mp-2026-10-22-013}

- Mainstream converters dropped color and highlighting cues, turning singled-out policy requirements into confident misses. [source-2026-10-22-013] — Qualification: The result concerns visual emphasis in this corpus, not every document or requirement.

## 14. Agents Can Be Mapped Without Running New Rollouts {#mp-2026-10-22-014}

- Story ID: `mp-2026-10-22-014`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-014/agents-can-be-mapped-without-running-new-rollouts

**Dek:** Centered action likelihoods on shared trajectories locate model-and-harness combinations by what they would do.

Agent Behavior Maps compare action log-likelihoods on the same reference trajectories instead of launching each target agent in a fresh environment. Across eight cross-scored models, distances followed measured trajectory divergence in rank order. Across 100 language models and four environments, ALFWorld-derived coordinates predicted performance and exposed behavioral similarities that persisted after controlling for success rate and model size. The guarantees depend on explicit assumptions, and the map is not a substitute for deployment testing.

### Why it matters {#why-it-matters-mp-2026-10-22-014}

Centered action likelihoods on shared trajectories locate model-and-harness combinations by what they would do.

### Limits and context {#limitations-mp-2026-10-22-014}

- The guarantees depend on explicit assumptions, and the map is not a substitute for deployment testing.

### Claims and sources {#claims-mp-2026-10-22-014}

- Centered action likelihoods on shared trajectories locate model-and-harness combinations by what they would do. [source-2026-10-22-014] — Qualification: The guarantees depend on explicit assumptions, and the map is not a substitute for deployment testing.

## 15. The Citation Exists. It Still Supports the Wrong Claim. {#mp-2026-10-22-026}

- Story ID: `mp-2026-10-22-026`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-026/the-citation-exists-it-still-supports-the-wrong-claim

**Dek:** Biomedical agents almost always produced real PubMed identifiers in closed-book mode, but almost none were relevant to the claim.

A controlled SciFact study compared closed-book answers, single-shot PubMed retrieval, a query-writing agent and access to gold abstracts. Of 335 closed-book citations, 327 PubMed identifiers existed, yet an independent judge found only three relevant to the claim. Letting the agent write queries tripled gold-paper recall and reached 62 percent verdict accuracy; gold abstracts lifted the same models to 79 to 86 percent, pointing to retrieval as the main bottleneck. The realistic-harness accuracy differences were not statistically significant at the sample size.

### Why it matters {#why-it-matters-mp-2026-10-22-026}

Biomedical agents almost always produced real PubMed identifiers in closed-book mode, but almost none were relevant to the claim.

### Limits and context {#limitations-mp-2026-10-22-026}

- Of 335 closed-book citations, 327 PubMed identifiers existed, yet an independent judge found only three relevant to the claim.
- The realistic-harness accuracy differences were not statistically significant at the sample size.

### Claims and sources {#claims-mp-2026-10-22-026}

- Biomedical agents almost always produced real PubMed identifiers in closed-book mode, but almost none were relevant to the claim. [source-2026-10-22-015] — Qualification: Of 335 closed-book citations, 327 PubMed identifiers existed, yet an independent judge found only three relevant to the claim.

## 16. Legal Models Still Choose When the Authority Is Missing {#mp-2026-10-22-027}

- Story ID: `mp-2026-10-22-027`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-027/legal-models-still-choose-when-the-authority-is-missing

**Dek:** Generative systems selected a wrong holding in 38 to 86 percent of abstention cases; a decision model cut errors but did not eliminate them.

Researchers replaced the correct holding in 1,670 of 5,314 CaseHOLD items and asked models to recognize that none of the options applied. Generative language models still chose an answer in 38 to 86 percent of those cases, while the most cautious probability-based decision model did so in 30 percent. On ordinary CaseHOLD, one decision model matched the best generative system at a reported one fifty-fifth of the cost. Agreement filtering halved that model's wrong citations. The benchmark is a controlled legal-authority task, not legal advice or field validation.

### Why it matters {#why-it-matters-mp-2026-10-22-027}

Generative systems selected a wrong holding in 38 to 86 percent of abstention cases; a decision model cut errors but did not eliminate them.

### Limits and context {#limitations-mp-2026-10-22-027}

- The benchmark is a controlled legal-authority task, not legal advice or field validation.

### Claims and sources {#claims-mp-2026-10-22-027}

- Generative systems selected a wrong holding in 38 to 86 percent of abstention cases; a decision model cut errors but did not eliminate them. [source-2026-10-22-016] — Qualification: The benchmark is a controlled legal-authority task, not legal advice or field validation.

## 17. One Silicon Surface Sees the Full Mid-Wave Infrared Band {#mp-2026-10-22-015}

- Story ID: `mp-2026-10-22-015`
- Type: `ticker`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-015/one-silicon-surface-sees-the-full-mid-wave-infrared-band

**Dek:** A two-centimeter hyperboloid metalens plus computation recovered images across 3 to 5 microns after adaptation from four captures.

The single-layer silicon metasurface leaves chromatic dispersion for a reconstruction model to untangle rather than engineering a complex achromatic profile. The authors report broadband imaging across the full mid-wave infrared band after adapting on four captures. This is a laboratory imaging result, not a deployed sensor claim.

### Why it matters {#why-it-matters-mp-2026-10-22-015}

A two-centimeter hyperboloid metalens plus computation recovered images across 3 to 5 microns after adaptation from four captures.

### Limits and context {#limitations-mp-2026-10-22-015}

- This is a laboratory imaging result, not a deployed sensor claim.

### Claims and sources {#claims-mp-2026-10-22-015}

- A two-centimeter hyperboloid metalens plus computation recovered images across 3 to 5 microns after adaptation from four captures. [source-2026-10-22-017] — Qualification: This is a laboratory imaging result, not a deployed sensor claim.

## 18. A Neutron Star Collision Builds a Star Within a Star {#mp-2026-10-22-016}

- Story ID: `mp-2026-10-22-016`
- Type: `ticker`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-016/a-neutron-star-collision-builds-a-star-within-a-star

**Dek:** A three-dimensional impact simulation followed by stellar evolution produced a bound Thorne-Żytkow object and a possible chemical signature.

The simulated neutron star settles inside a ten-solar-mass stellar envelope. Collision heating triggers broad nuclear burning before convection carries carbon-nitrogen-cycle material to the surface. The work connects formation and long-term evolution in simulation; it does not report an observed Thorne-Żytkow object.

### Why it matters {#why-it-matters-mp-2026-10-22-016}

A three-dimensional impact simulation followed by stellar evolution produced a bound Thorne-Żytkow object and a possible chemical signature.

### Limits and context {#limitations-mp-2026-10-22-016}

- The work connects formation and long-term evolution in simulation; it does not report an observed Thorne-Żytkow object.

### Claims and sources {#claims-mp-2026-10-22-016}

- A three-dimensional impact simulation followed by stellar evolution produced a bound Thorne-Żytkow object and a possible chemical signature. [source-2026-10-22-018] — Qualification: The work connects formation and long-term evolution in simulation; it does not report an observed Thorne-Żytkow object.

## 19. Reasoning Monitorability Becomes a Training Target {#mp-2026-10-22-017}

- Story ID: `mp-2026-10-22-017`
- Type: `ticker`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-017/reasoning-monitorability-becomes-a-training-target

**Dek:** An unsupervised mutual-information objective recovered much of the monitoring signal lost in three degraded chain-of-thought model organisms.

Post-training strengthened the connection between internal reasoning text and model outputs, improving judged monitorability and prediction of unsafe behavior in the reported tests. The result suggests monitorability can be optimized, but it does not make chain-of-thought a complete or tamper-proof oversight channel.

### Why it matters {#why-it-matters-mp-2026-10-22-017}

An unsupervised mutual-information objective recovered much of the monitoring signal lost in three degraded chain-of-thought model organisms.

### Limits and context {#limitations-mp-2026-10-22-017}

- The result suggests monitorability can be optimized, but it does not make chain-of-thought a complete or tamper-proof oversight channel.

### Claims and sources {#claims-mp-2026-10-22-017}

- An unsupervised mutual-information objective recovered much of the monitoring signal lost in three degraded chain-of-thought model organisms. [source-2026-10-22-019] — Qualification: The result suggests monitorability can be optimized, but it does not make chain-of-thought a complete or tamper-proof oversight channel.

## 20. Three Monitors Turn Silent Faults Into Diagnoses {#mp-2026-10-22-018}

- Story ID: `mp-2026-10-22-018`
- Type: `ticker`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-018/three-monitors-turn-silent-faults-into-diagnoses

**Dek:** COMET combines signal integrity, physics consistency and competence-envelope checks for human-supervised safety systems.

On compressed-air leak data with simulated distribution shifts and sensor faults, the bare AI failed silently and confidently. COMET reduced trusted wrong answers to zero on shifted no-leak data and three of four sensor faults at about a one-percent nominal false-alarm rate. Semantic-shift misses remained, marking a boundary for passive monitoring.

### Why it matters {#why-it-matters-mp-2026-10-22-018}

COMET combines signal integrity, physics consistency and competence-envelope checks for human-supervised safety systems.

### Limits and context {#limitations-mp-2026-10-22-018}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-10-22-018}

- COMET combines signal integrity, physics consistency and competence-envelope checks for human-supervised safety systems. [source-2026-10-22-020]

## 21. The Pouring Robot Estimates What the Camera Cannot See {#mp-2026-10-22-019}

- Story ID: `mp-2026-10-22-019`
- Type: `ticker`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-019/the-pouring-robot-estimates-what-the-camera-cannot-see

**Dek:** FlowVLA tracks transferred mass from visual, joint, wrist-force and recurrent histories, then predicts residual outflow.

In coffee-pot and water tests, the controller reported 13.1 grams mean absolute error across goals; 53.7 percent of attempts both completed and landed within ten percent of the requested mass. The same model interpolated to held-out 150- and 250-gram targets. Results are task-specific and still leave substantial error.

### Why it matters {#why-it-matters-mp-2026-10-22-019}

FlowVLA tracks transferred mass from visual, joint, wrist-force and recurrent histories, then predicts residual outflow.

### Limits and context {#limitations-mp-2026-10-22-019}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-10-22-019}

- FlowVLA tracks transferred mass from visual, joint, wrist-force and recurrent histories, then predicts residual outflow. [source-2026-10-22-021]

## 22. Space Drums 2.0 {#mp-2026-10-22-020}

- Story ID: `mp-2026-10-22-020`
- Type: `invention_desk`
- Classification: `editorial`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-020/space-drums-2-0

**Dek:** Moves motion tracking onto two ESP32-S3 drumsticks, adds haptic hits and a two-pedal hub, and sends velocity-sensitive drum events to open PC or Android playback software.

Moves motion tracking onto two ESP32-S3 drumsticks, adds haptic hits and a two-pedal hub, and sends velocity-sensitive drum events to open PC or Android playback software.

### Why it matters {#why-it-matters-mp-2026-10-22-020}

An independent builder is turning an improbable idea into a working project.

### Limits and context {#limitations-mp-2026-10-22-020}

- A Desk Pick is an editorial selection, not a product endorsement.

### Claims and sources {#claims-mp-2026-10-22-020}

- This Invention Desk entry makes no independently sourced news claim.

## 23. 3RIC {#mp-2026-10-22-021}

- Story ID: `mp-2026-10-22-021`
- Type: `invention_desk`
- Classification: `editorial`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-021/3ric

**Dek:** Pairs a chip-level 65C02 computer with custom hardware, video, sound, floppy, and microSD support with a cycle-honest emulator and browser assembler built around the same machine.

Pairs a chip-level 65C02 computer with custom hardware, video, sound, floppy, and microSD support with a cycle-honest emulator and browser assembler built around the same machine.

### Why it matters {#why-it-matters-mp-2026-10-22-021}

An independent builder is turning an improbable idea into a working project.

### Limits and context {#limitations-mp-2026-10-22-021}

- A Desk Pick is an editorial selection, not a product endorsement.

### Claims and sources {#claims-mp-2026-10-22-021}

- This Invention Desk entry makes no independently sourced news claim.

## 24. M3-CRETE {#mp-2026-10-22-022}

- Story ID: `mp-2026-10-22-022`
- Type: `invention_desk`
- Classification: `editorial`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-022/m3-crete

**Dek:** Publishes a pallet-scale cartesian motion reference, CAD, BOM, and low-voltage controls for supervised cementitious-extrusion research while explicitly excluding certified structural use.

Publishes a pallet-scale cartesian motion reference, CAD, BOM, and low-voltage controls for supervised cementitious-extrusion research while explicitly excluding certified structural use.

### Why it matters {#why-it-matters-mp-2026-10-22-022}

An independent builder is turning an improbable idea into a working project.

### Limits and context {#limitations-mp-2026-10-22-022}

- A Desk Pick is an editorial selection, not a product endorsement.

### Claims and sources {#claims-mp-2026-10-22-022}

- This Invention Desk entry makes no independently sourced news claim.

## 25. DeskNode {#mp-2026-10-22-023}

- Story ID: `mp-2026-10-22-023`
- Type: `invention_desk`
- Classification: `editorial`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-023/desknode

**Dek:** Turns a bare ESP32-S3 touchscreen into a USB-connected desk instrument for PC activity and optional room temperature, humidity, and light measurements.

Turns a bare ESP32-S3 touchscreen into a USB-connected desk instrument for PC activity and optional room temperature, humidity, and light measurements.

### Why it matters {#why-it-matters-mp-2026-10-22-023}

An independent builder is turning an improbable idea into a working project.

### Limits and context {#limitations-mp-2026-10-22-023}

- A Desk Pick is an editorial selection, not a product endorsement.

### Claims and sources {#claims-mp-2026-10-22-023}

- This Invention Desk entry makes no independently sourced news claim.

## 26. The First Paid Slot {#mp-2026-10-22-024}

- Story ID: `mp-2026-10-22-024`
- Type: `invention_desk`
- Classification: `house_example`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-024/the-first-paid-slot

**Dek:** A transparent preview of paid placement with one verified link and no claim of endorsement.

A transparent preview of paid placement with one verified link and no claim of endorsement.

House example - no advertiser paid. Payment will buy placement, never endorsement.

### Why it matters {#why-it-matters-mp-2026-10-22-024}

This placement explains how builders can appear in The Invention Desk without purchasing editorial endorsement.

### Limits and context {#limitations-mp-2026-10-22-024}

- House example - no advertiser paid. Payment will buy placement, never endorsement.

### Claims and sources {#claims-mp-2026-10-22-024}

- This Invention Desk entry makes no independently sourced news claim.

## 27. Put Your Project on the Desk {#mp-2026-10-22-025}

- Story ID: `mp-2026-10-22-025`
- Type: `invention_desk`
- Classification: `house_example`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-22-025/put-your-project-on-the-desk

**Dek:** One manually reviewed placement stays active for seven days and remains separate from Desk Picks.

One manually reviewed placement stays active for seven days and remains separate from Desk Picks.

Manual intake only. Payment buys placement, never endorsement, and every submission is reviewed.

### Why it matters {#why-it-matters-mp-2026-10-22-025}

This placement explains how builders can appear in The Invention Desk without purchasing editorial endorsement.

### Limits and context {#limitations-mp-2026-10-22-025}

- Manual intake only. Payment buys placement, never endorsement, and every submission is reviewed.

### Claims and sources {#claims-mp-2026-10-22-025}

- This Invention Desk entry makes no independently sourced news claim.

## Normalized sources

- **source-2026-10-22-001:** [arXiv preprint 2610.29986, original v1](https://arxiv.org/abs/2610.29986v1) — arXiv; primary_research
- **source-2026-10-22-002:** [arXiv preprint 2610.29869, original v1](https://arxiv.org/abs/2610.29869v1) — arXiv; primary_research
- **source-2026-10-22-003:** [arXiv preprint 2610.30136, original v1](https://arxiv.org/abs/2610.30136v1) — arXiv; primary_research
- **source-2026-10-22-004:** [arXiv preprint 2610.30129, original v1](https://arxiv.org/abs/2610.30129v1) — arXiv; primary_research
- **source-2026-10-22-005:** [arXiv preprint 2610.30125, original v1](https://arxiv.org/abs/2610.30125v1) — arXiv; primary_research
- **source-2026-10-22-006:** [arXiv preprint 2610.30117, original v1](https://arxiv.org/abs/2610.30117v1) — arXiv; primary_research
- **source-2026-10-22-007:** [arXiv preprint 2610.30105, original v1](https://arxiv.org/abs/2610.30105v1) — arXiv; primary_research
- **source-2026-10-22-008:** [arXiv preprint 2610.30104, original v1](https://arxiv.org/abs/2610.30104v1) — arXiv; primary_research
- **source-2026-10-22-009:** [arXiv preprint 2610.30097, original v1](https://arxiv.org/abs/2610.30097v1) — arXiv; primary_research
- **source-2026-10-22-010:** [arXiv preprint 2610.30080, original v1](https://arxiv.org/abs/2610.30080v1) — arXiv; primary_research
- **source-2026-10-22-011:** [arXiv preprint 2610.30061, original v1](https://arxiv.org/abs/2610.30061v1) — arXiv; primary_research
- **source-2026-10-22-012:** [arXiv preprint 2610.30035, original v1](https://arxiv.org/abs/2610.30035v1) — arXiv; primary_research
- **source-2026-10-22-013:** [arXiv preprint 2610.30026, original v1](https://arxiv.org/abs/2610.30026v1) — arXiv; primary_research
- **source-2026-10-22-014:** [arXiv preprint 2610.30014, original v1](https://arxiv.org/abs/2610.30014v1) — arXiv; primary_research
- **source-2026-10-22-015:** [arXiv preprint 2610.29812, original v1](https://arxiv.org/abs/2610.29812v1) — arXiv; primary_research
- **source-2026-10-22-016:** [arXiv preprint 2610.29961, original v1](https://arxiv.org/abs/2610.29961v1) — arXiv; primary_research
- **source-2026-10-22-017:** [arXiv preprint 2610.29841, original v1](https://arxiv.org/abs/2610.29841v1) — arXiv; primary_research
- **source-2026-10-22-018:** [arXiv preprint 2610.29833, original v1](https://arxiv.org/abs/2610.29833v1) — arXiv; primary_research
- **source-2026-10-22-019:** [arXiv preprint 2610.29734, original v1](https://arxiv.org/abs/2610.29734v1) — arXiv; primary_research
- **source-2026-10-22-020:** [arXiv preprint 2610.29816, original v1](https://arxiv.org/abs/2610.29816v1) — arXiv; primary_research
- **source-2026-10-22-021:** [arXiv preprint 2610.29692, original v1](https://arxiv.org/abs/2610.29692v1) — arXiv; primary_research