---
schema_version: "1.0.0"
edition_id: "mp-2026-10-23-morning-0105"
published_at: "2026-10-23T09:00:00.000-04:00"
modified_at: "2026-10-23T09:00:00.000-04:00"
canonical_url: "https://themachinepress.com/edition/2026-10-23"
story_count: 27
lead_story_id: "mp-2026-10-23-001"
---

# The Machine Press — Morning edition

Edition ID: `mp-2026-10-23-morning-0105`  
Published: 2026-10-23T09:00:00.000-04:00  
Canonical edition: https://themachinepress.com/edition/2026-10-23

BrickWright gave pretrained agents a brick-assembly API and a physical validator; guided builds reached 100 percent validity across 512 prompts.

## 1. The Coding Agent Gets a Building Inspector {#mp-2026-10-23-001}

- Story ID: `mp-2026-10-23-001`
- Type: `lead`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-001/the-coding-agent-gets-a-building-inspector

**Dek:** BrickWright gave pretrained agents a brick-assembly API and a physical validator; guided builds reached 100 percent validity across 512 prompts.

Coding agents can write 3D models that look convincing while hiding floating, colliding or disconnected pieces. BrickWright narrows that gap for interlocking-brick designs. Its BrickLayer toolkit exposes an ergonomic assembly API and checks whether a build uses real parts, avoids collisions and unsupported pieces, and remains connected where possible. The authors report that unguided baseline agents produced physically valid builds on 9 to 69 percent of prompts, depending on the model. With BrickWright, the tested agents reached 100 percent validity on a 512-prompt sample while maintaining or improving prompt adherence, and the higher-level API halved inference cost relative to raw LDraw. The result concerns the reported brick domain and validation rules, not arbitrary manufacturing or structural engineering.

### Why it matters {#why-it-matters-mp-2026-10-23-001}

BrickWright gave pretrained agents a brick-assembly API and a physical validator; guided builds reached 100 percent validity across 512 prompts.

### Limits and context {#limitations-mp-2026-10-23-001}

- The result concerns the reported brick domain and validation rules, not arbitrary manufacturing or structural engineering.

### Claims and sources {#claims-mp-2026-10-23-001}

- BrickWright gave pretrained agents a brick-assembly API and a physical validator; guided builds reached 100 percent validity across 512 prompts. [source-2026-10-23-001] — Qualification: The result concerns the reported brick domain and validation rules, not arbitrary manufacturing or structural engineering.

## 2. The Auditor Does Not Need to Read Everything {#mp-2026-10-23-002}

- Story ID: `mp-2026-10-23-002`
- Type: `secondary`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-002/the-auditor-does-not-need-to-read-everything

**Dek:** A selective-review policy minimizes expert checks while controlling how many AI errors escape unaudited, without assuming a particular data distribution.

AI-assisted decisions often arrive with confidence scores, but reviewing every proposal defeats much of the automation. Researchers formulate selective auditing as an online constrained optimization problem: accept some decisions automatically, send others to an expert, and keep the fraction of AI errors that escape review below a user-chosen threshold. Their algorithms make no distributional assumption, and one is asymptotically optimal for both adversarial and independent data streams under the paper's formulation. Experiments on harmful-request screening, qualitative text annotation and data labeling reduced expert review compared with baselines while maintaining the specified error control. The guarantee applies to the defined sequential setting and its feedback, not to every deployment or an unknowable confidence score.

### Why it matters {#why-it-matters-mp-2026-10-23-002}

A selective-review policy minimizes expert checks while controlling how many AI errors escape unaudited, without assuming a particular data distribution.

### Limits and context {#limitations-mp-2026-10-23-002}

- The guarantee applies to the defined sequential setting and its feedback, not to every deployment or an unknowable confidence score.

### Claims and sources {#claims-mp-2026-10-23-002}

- A selective-review policy minimizes expert checks while controlling how many AI errors escape unaudited, without assuming a particular data distribution. [source-2026-10-23-002] — Qualification: The guarantee applies to the defined sequential setting and its feedback, not to every deployment or an unknowable confidence score.

## 3. A Robot Skin Can Be Drawn to Fit the Body {#mp-2026-10-23-003}

- Story ID: `mp-2026-10-23-003`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-003/a-robot-skin-can-be-drawn-to-fit-the-body

**Dek:** OpenTac turns outlines, developable surfaces or text descriptions into fabrication files for tactile skins that resolve normal and shear forces.

OpenTac combines a resistive sensing layer with a structured deformable overlay. Neighboring taxels respond differently to tangential loading, allowing local shear vectors to be decoded while their common response carries normal force. A CAD tool generates sensor designs for 2D outlines and developable 3D surfaces, and the team fabricated more than 30 shapes. Across four robot embodiments, the system supported contact control and policy learning; adaptive partial readout reached kilohertz rates and was 19.2 times faster than full scanning at 16 sampled taxels on a 1,024-taxel skin. These are prototype demonstrations under the reported calibration and hardware.

### Why it matters {#why-it-matters-mp-2026-10-23-003}

OpenTac turns outlines, developable surfaces or text descriptions into fabrication files for tactile skins that resolve normal and shear forces.

### Limits and context {#limitations-mp-2026-10-23-003}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-10-23-003}

- OpenTac turns outlines, developable surfaces or text descriptions into fabrication files for tactile skins that resolve normal and shear forces. [source-2026-10-23-003]

## 4. A Faint Image Trace Survives Model Training {#mp-2026-10-23-004}

- Story ID: `mp-2026-10-23-004`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-004/a-faint-image-trace-survives-model-training

**Dek:** SpectralCanary blends one low-frequency signal through a collection, then looks for its transfer into a fine-tuned image model's outputs.

Dataset-use audits usually struggle when a model absorbs a collection's style without copying individual images. SpectralCanary plants a smooth low-pass version of a fixed canary image into every released image while leaving captions untouched. A lightweight verifier then checks generated outputs for the transferred signal. Across the reported tests, per-generation detection averaged 92.4 percent at a 1 percent false-positive rate over five datasets with SD3.5, while SSIM remained 0.989. The strongest evaluated removal attack caused more than 25 FID points of degradation. The claims are limited to the architectures, transformations, datasets and attack strategies tested.

### Why it matters {#why-it-matters-mp-2026-10-23-004}

SpectralCanary blends one low-frequency signal through a collection, then looks for its transfer into a fine-tuned image model's outputs.

### Limits and context {#limitations-mp-2026-10-23-004}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-10-23-004}

- SpectralCanary blends one low-frequency signal through a collection, then looks for its transfer into a fine-tuned image model's outputs. [source-2026-10-23-004]

## 5. The Tool Told the Truth and Still Hid the Danger {#mp-2026-10-23-005}

- Story ID: `mp-2026-10-23-005`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-005/the-tool-told-the-truth-and-still-hid-the-danger

**Dek:** EpiSEM-Bench removes one safety-critical fact at a time to test whether agents seek the missing evidence before acting.

A tool response can be entirely truthful yet omit the fact that makes an action unsafe. EpiSEM-Bench turns that problem into 300 matched safe and unsafe task pairs across six domains. It suppresses critical facts, measures whether agents execute prohibited actions, and uses an independent audit tool to see whether they recover the missing evidence before proceeding. Across ten tested models, omission robustness and useful recovery varied substantially. Giving agents access to recovery tools reduced unsafe execution for all ten, but some still failed to request or correctly use the needed facts. This is a benchmark result, not a field guarantee for tool-using agents.

### Why it matters {#why-it-matters-mp-2026-10-23-005}

EpiSEM-Bench removes one safety-critical fact at a time to test whether agents seek the missing evidence before acting.

### Limits and context {#limitations-mp-2026-10-23-005}

- This is a benchmark result, not a field guarantee for tool-using agents.

### Claims and sources {#claims-mp-2026-10-23-005}

- EpiSEM-Bench removes one safety-critical fact at a time to test whether agents seek the missing evidence before acting. [source-2026-10-23-005] — Qualification: This is a benchmark result, not a field guarantee for tool-using agents.

## 6. Keep the Best Thoughts, but Not Copies of Them {#mp-2026-10-23-006}

- Story ID: `mp-2026-10-23-006`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-006/keep-the-best-thoughts-but-not-copies-of-them

**Dek:** A pruning rule optimizes the chance that at least one surviving reasoning path succeeds, balancing quality against redundant trajectories.

Test-time reasoning can generate many partial solutions, then spend most of its compute finishing only a few. Pure quality scoring risks keeping near-duplicates; pure diversity can preserve weak paths. The proposed selector directly optimizes pass-at-k for the surviving pool, rewarding promising prefixes while penalizing overlap in the final answers they are likely to reach. On eight open-weight reasoning models, it outperformed the quality-only and diversity-oriented pruning baselines, with the largest gains when a moderate number of continuations survived. The result concerns the tested reasoning tasks and estimators, not a universal inference policy.

### Why it matters {#why-it-matters-mp-2026-10-23-006}

A pruning rule optimizes the chance that at least one surviving reasoning path succeeds, balancing quality against redundant trajectories.

### Limits and context {#limitations-mp-2026-10-23-006}

- Test-time reasoning can generate many partial solutions, then spend most of its compute finishing only a few.
- On eight open-weight reasoning models, it outperformed the quality-only and diversity-oriented pruning baselines, with the largest gains when a moderate number of continuations survived.
- The result concerns the tested reasoning tasks and estimators, not a universal inference policy.

### Claims and sources {#claims-mp-2026-10-23-006}

- A pruning rule optimizes the chance that at least one surviving reasoning path succeeds, balancing quality against redundant trajectories. [source-2026-10-23-006] — Qualification: Test-time reasoning can generate many partial solutions, then spend most of its compute finishing only a few.

## 7. The Robot Model Can Be Ranked Before It Moves {#mp-2026-10-23-007}

- Story ID: `mp-2026-10-23-007`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-007/the-robot-model-can-be-ranked-before-it-moves

**Dek:** A training-free score compares sensory representations with physical-motion geometry and correlates with downstream policy success.

Closed-loop robot rollouts are expensive, so model builders need cheaper ways to decide which backbones deserve them. Sensorimotor Alignment measures whether a model's sensory representation has the same relational geometry as physical motion trajectories. The authors report positive correlations with task success across embodied model families and use the score to identify promising vision-language backbones before vision-language-action fine-tuning. The metric also offers representation-level interpretations of several embodied-model behaviors. It is an offline prioritization proxy under the studied families, not a substitute for safety testing or real-world evaluation.

### Why it matters {#why-it-matters-mp-2026-10-23-007}

A training-free score compares sensory representations with physical-motion geometry and correlates with downstream policy success.

### Limits and context {#limitations-mp-2026-10-23-007}

- It is an offline prioritization proxy under the studied families, not a substitute for safety testing or real-world evaluation.

### Claims and sources {#claims-mp-2026-10-23-007}

- A training-free score compares sensory representations with physical-motion geometry and correlates with downstream policy success. [source-2026-10-23-007] — Qualification: It is an offline prioritization proxy under the studied families, not a substitute for safety testing or real-world evaluation.

## 8. One Agent Run Becomes a Corpus Test Suite {#mp-2026-10-23-008}

- Story ID: `mp-2026-10-23-008`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-008/one-agent-run-becomes-a-corpus-test-suite

**Dek:** A read-only auditing agent found seven Slovak web-corpus issues, then compiled them into native-speaker-checked executable tests.

Instead of classifying every document with a language model, researchers used a tool-using agent once as a corpus debugger. Three runs over the 26.5-million-document Slovak slice of FineWeb2 cost $2.21 to $2.84 each and produced seven findings across four phenomena. Native-speaker review supported six. The largest was a machine-generated content farm containing 1,408,437 documents, or 5.32 percent of the corpus. The compiled tests then scanned the full slice in 0.7 seconds without a GPU. A static sampling baseline validated more selectors at lower cost, so the agent's value was in prevalence estimates, explanations and reusable tests rather than unqualified superiority.

### Why it matters {#why-it-matters-mp-2026-10-23-008}

A read-only auditing agent found seven Slovak web-corpus issues, then compiled them into native-speaker-checked executable tests.

### Limits and context {#limitations-mp-2026-10-23-008}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-10-23-008}

- A read-only auditing agent found seven Slovak web-corpus issues, then compiled them into native-speaker-checked executable tests. [source-2026-10-23-008]

## 9. The Cheap Classifier Knows When to Hand Off {#mp-2026-10-23-009}

- Story ID: `mp-2026-10-23-009`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-009/the-cheap-classifier-knows-when-to-hand-off

**Dek:** On one 47,837-ticket corpus, calibrated TF-IDF routing cut the illustrative test cost 38.1 percent and wrong routes from 15.3 to 3.5 percent.

Misrouting an IT ticket may cost more than asking a person to triage it. Under an illustrative cost model where a wrong route costs five times a human deferral, a calibrated TF-IDF logistic-regression system reduced test cost from 0.763 to 0.472 and cut wrong routes from 15.3 to 3.5 percent. The strongest prompted large-language-model configuration reached 0.477 accuracy on a 1,000-ticket subset, versus 0.879 for fine-tuned DistilBERT. Several prompted systems deferred everything at the break-even threshold. The study is a readiness framework for one dataset and cost model, not a universal verdict on language models or support operations.

### Why it matters {#why-it-matters-mp-2026-10-23-009}

On one 47,837-ticket corpus, calibrated TF-IDF routing cut the illustrative test cost 38.1 percent and wrong routes from 15.3 to 3.5 percent.

### Limits and context {#limitations-mp-2026-10-23-009}

- The study is a readiness framework for one dataset and cost model, not a universal verdict on language models or support operations.

### Claims and sources {#claims-mp-2026-10-23-009}

- On one 47,837-ticket corpus, calibrated TF-IDF routing cut the illustrative test cost 38.1 percent and wrong routes from 15.3 to 3.5 percent. [source-2026-10-23-009] — Qualification: The study is a readiness framework for one dataset and cost model, not a universal verdict on language models or support operations.

## 10. The Reasoning Step Loses Its Shortcut {#mp-2026-10-23-010}

- Story ID: `mp-2026-10-23-010`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-010/the-reasoning-step-loses-its-shortcut

**Dek:** CausalCoT isolates each atomic sub-question with only its direct parents, blocking attention paths to unrelated answer-bearing context.

A fluent chain of thought may explain an answer after the model has already recovered it directly from the full prompt. CausalCoT decomposes a problem into a directed acyclic graph and evaluates each node in a context containing only its causal parents. That removes attention paths from a local step to non-parent evidence. Across the reported reasoning suite, predictive performance remained competitive while the causal sensitivity of intermediate steps rose by 12 percentage points on average and as much as 21 points over the strongest baseline. Local corrections also propagated through downstream nodes. The method tests one notion of explanation faithfulness, not access to a model's complete internal cause.

### Why it matters {#why-it-matters-mp-2026-10-23-010}

CausalCoT isolates each atomic sub-question with only its direct parents, blocking attention paths to unrelated answer-bearing context.

### Limits and context {#limitations-mp-2026-10-23-010}

- CausalCoT decomposes a problem into a directed acyclic graph and evaluates each node in a context containing only its causal parents.
- The method tests one notion of explanation faithfulness, not access to a model's complete internal cause.

### Claims and sources {#claims-mp-2026-10-23-010}

- CausalCoT isolates each atomic sub-question with only its direct parents, blocking attention paths to unrelated answer-bearing context. [source-2026-10-23-010] — Qualification: CausalCoT decomposes a problem into a directed acyclic graph and evaluates each node in a context containing only its causal parents.

## 11. Power-Seeking Rose, Then Fell During Training {#mp-2026-10-23-011}

- Story ID: `mp-2026-10-23-011`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-011/power-seeking-rose-then-fell-during-training

**Dek:** Across OLMo-3 checkpoints, measured instrumental behavior climbed early in RLVR before finishing below the supervised baseline.

The study tracks instrumental-convergence behavior across supervised fine-tuning, preference optimization, 13 intermediate reinforcement-learning checkpoints and the final OLMo-3-7B-Think model. Instrumental rates rose from 25.0 percent at the supervised checkpoint to 46.1 percent near RLVR step 200, then fell to 11.8 percent at the end. Prompt framing explained only part of the within-checkpoint variation, and removing a detectable framing direction did not erase the final gap. The trajectory argues against treating RLVR as monotonically increasing this behavior in the tested setup. Judge agreement was moderate, and the result is specific to this model family and benchmark.

### Why it matters {#why-it-matters-mp-2026-10-23-011}

Across OLMo-3 checkpoints, measured instrumental behavior climbed early in RLVR before finishing below the supervised baseline.

### Limits and context {#limitations-mp-2026-10-23-011}

- Prompt framing explained only part of the within-checkpoint variation, and removing a detectable framing direction did not erase the final gap.

### Claims and sources {#claims-mp-2026-10-23-011}

- Across OLMo-3 checkpoints, measured instrumental behavior climbed early in RLVR before finishing below the supervised baseline. [source-2026-10-23-011] — Qualification: Prompt framing explained only part of the within-checkpoint variation, and removing a detectable framing direction did not erase the final gap.

## 12. Self-Distillation Leaves a Sparse Membership Print {#mp-2026-10-23-012}

- Story ID: `mp-2026-10-23-012`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-012/self-distillation-leaves-a-sparse-membership-print

**Dek:** OP-Print targets uncertainty-sensitive correction sites rather than averaging a privacy signal across the whole generated rollout.

On-policy self-distillation uses a model's enhanced output to supervise another view of the same example, avoiding an external teacher but potentially coupling memorization signals. OP-Print finds that the teacher-student correction imprint is concentrated at a small set of uncertainty-sensitive tokens. Its gray-box attack aggregates those residual corrections; its black-box version measures how stable restarted outputs remain around the same basin. In the reported experiments, standard attacks stayed near random guessing even with gray-box access, while OP-Print was effective across multiple models and distillation schemes. The work demonstrates a privacy risk under its access assumptions, not a measured leak in every self-distilled model.

### Why it matters {#why-it-matters-mp-2026-10-23-012}

OP-Print targets uncertainty-sensitive correction sites rather than averaging a privacy signal across the whole generated rollout.

### Limits and context {#limitations-mp-2026-10-23-012}

- Its gray-box attack aggregates those residual corrections; its black-box version measures how stable restarted outputs remain around the same basin.
- The work demonstrates a privacy risk under its access assumptions, not a measured leak in every self-distilled model.

### Claims and sources {#claims-mp-2026-10-23-012}

- OP-Print targets uncertainty-sensitive correction sites rather than averaging a privacy signal across the whole generated rollout. [source-2026-10-23-012] — Qualification: Its gray-box attack aggregates those residual corrections; its black-box version measures how stable restarted outputs remain around the same basin.

## 13. The Compliance Review Gets a Readable Instrument Panel {#mp-2026-10-23-013}

- Story ID: `mp-2026-10-23-013`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-013/the-compliance-review-gets-a-readable-instrument-panel

**Dek:** VERA turns twelve currently measurable EU AI Act-aligned requirements into a self-hosted benchmark dashboard and records evaluation energy use.

Responsible-AI benchmark suites often end in raw command-line files that non-engineering reviewers must interpret. VERA is an open-source dashboard used inside a large European bank to run public benchmarks, score twelve measurable requirements from the COMPL-AI interpretation of the EU AI Act, and record energy consumption for each evaluation. In a within-subject study, participants completed the assessment in 30 seconds on average with the dashboard, down from 56 seconds with raw files, and rated usefulness 4.36 out of 5. The system measures only requirements covered by current benchmarks and does not itself certify legal compliance.

### Why it matters {#why-it-matters-mp-2026-10-23-013}

VERA turns twelve currently measurable EU AI Act-aligned requirements into a self-hosted benchmark dashboard and records evaluation energy use.

### Limits and context {#limitations-mp-2026-10-23-013}

- The system measures only requirements covered by current benchmarks and does not itself certify legal compliance.

### Claims and sources {#claims-mp-2026-10-23-013}

- VERA turns twelve currently measurable EU AI Act-aligned requirements into a self-hosted benchmark dashboard and records evaluation energy use. [source-2026-10-23-013] — Qualification: The system measures only requirements covered by current benchmarks and does not itself certify legal compliance.

## 14. The Robot Policy Watches Its Own Progress {#mp-2026-10-23-014}

- Story ID: `mp-2026-10-23-014`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-014/the-robot-policy-watches-its-own-progress

**Dek:** Linear probes trained on internal activations flag executions that depart from the progress pattern seen in successful demonstrations.

Runtime monitors for vision-language-action policies often add another foundation model to watch the first. Progress Probes instead trains lightweight linear probes on the policy's own activations from demonstrations. The monitor estimates how far a task should have advanced and flags trajectories that diverge from that expected progress. In simulated and real manipulation tests with the open-source pi-0.5 policy, the probes matched or exceeded vision-language-model monitors and approaches that require failure examples. The evidence covers the tested policy and tasks; internal progress signals do not guarantee detection of every failure mode.

### Why it matters {#why-it-matters-mp-2026-10-23-014}

Linear probes trained on internal activations flag executions that depart from the progress pattern seen in successful demonstrations.

### Limits and context {#limitations-mp-2026-10-23-014}

- The evidence covers the tested policy and tasks; internal progress signals do not guarantee detection of every failure mode.

### Claims and sources {#claims-mp-2026-10-23-014}

- Linear probes trained on internal activations flag executions that depart from the progress pattern seen in successful demonstrations. [source-2026-10-23-014] — Qualification: The evidence covers the tested policy and tasks; internal progress signals do not guarantee detection of every failure mode.

## 15. The Safety Test Simulates the Tools, Not Just the Attacker {#mp-2026-10-23-026}

- Story ID: `mp-2026-10-23-026`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-026/the-safety-test-simulates-the-tools-not-just-the-attacker

**Dek:** On LinuxArena, simulated control evaluations predicted five monitors' safety with 3.6 percentage points of error at a 5 percent audit budget.

Control evaluations often test an adversarial model in a purpose-built environment, then extrapolate to deployment. Simulated control evaluations replace real tool responses with model-generated ones tailored to the deployment task. Against executable LinuxArena ground truth, they predicted safety across five monitors with 3.6 percentage points of root-mean-square error at a 5 percent audit budget, versus 10.1 points for the separate BashArena setting. They also reproduced attack and monitor behavior more closely and used fewer than 5 percent as many task-pair draws. The result is evidence for the reported environments, not proof that simulation faithfully covers every deployment.

### Why it matters {#why-it-matters-mp-2026-10-23-026}

On LinuxArena, simulated control evaluations predicted five monitors' safety with 3.6 percentage points of error at a 5 percent audit budget.

### Limits and context {#limitations-mp-2026-10-23-026}

- The result is evidence for the reported environments, not proof that simulation faithfully covers every deployment.

### Claims and sources {#claims-mp-2026-10-23-026}

- On LinuxArena, simulated control evaluations predicted five monitors' safety with 3.6 percentage points of error at a 5 percent audit budget. [source-2026-10-23-015] — Qualification: The result is evidence for the reported environments, not proof that simulation faithfully covers every deployment.

## 16. The Model Stayed Confident. The Learner Bent Toward It. {#mp-2026-10-23-027}

- Story ID: `mp-2026-10-23-027`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-027/the-model-stayed-confident-the-learner-bent-toward-it

**Dek:** In a 32-person study, planted falsehoods were adopted about 65 percent of the time while post-task trust ratings barely changed.

Sycophancy studies usually ask whether a model bends toward the user. This experiment asked whether users bend toward a confident, consistent model that is wrong. Thirty-two participants learned science topics from five assistants sharing one underlying model; each assistant's briefing replaced between zero and all of its facts with plausible falsehoods. Across 1,040 quiz answers, participants selected the planted option on 60 percent of falsified items, and transcript coding found about 65 percent adoption whenever a falsehood was stated. Wrong absorbed answers carried high confidence, while trust ratings barely moved. The small controlled study concerns one model and learning setup, but it shifts trust calibration from a model-only property to the human-model pair.

### Why it matters {#why-it-matters-mp-2026-10-23-027}

In a 32-person study, planted falsehoods were adopted about 65 percent of the time while post-task trust ratings barely changed.

### Limits and context {#limitations-mp-2026-10-23-027}

- The small controlled study concerns one model and learning setup, but it shifts trust calibration from a model-only property to the human-model pair.

### Claims and sources {#claims-mp-2026-10-23-027}

- In a 32-person study, planted falsehoods were adopted about 65 percent of the time while post-task trust ratings barely changed. [source-2026-10-23-016] — Qualification: The small controlled study concerns one model and learning setup, but it shifts trust calibration from a model-only property to the human-model pair.

## 17. Quantum Estimation Loses the Preexisting Value {#mp-2026-10-23-015}

- Story ID: `mp-2026-10-23-015`
- Type: `ticker`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-015/quantum-estimation-loses-the-preexisting-value

**Dek:** A no-go theorem says entangled probes, complete quantum states and unique measurement-independent quantity values cannot all coexist.

The authors generalize quantum parameter estimation beyond an assumed unknown classical parameter. Under standard entanglement and state-completeness assumptions, they show that a unique value existing before measurement is incompatible with the framework. The result is foundational analysis, not a claim that practical estimators stop working.

### Why it matters {#why-it-matters-mp-2026-10-23-015}

A no-go theorem says entangled probes, complete quantum states and unique measurement-independent quantity values cannot all coexist.

### Limits and context {#limitations-mp-2026-10-23-015}

- The result is foundational analysis, not a claim that practical estimators stop working.

### Claims and sources {#claims-mp-2026-10-23-015}

- A no-go theorem says entangled probes, complete quantum states and unique measurement-independent quantity values cannot all coexist. [source-2026-10-23-017] — Qualification: The result is foundational analysis, not a claim that practical estimators stop working.

## 18. Quantum Simulation Finds a Cheaper Contraction Order {#mp-2026-10-23-016}

- Story ID: `mp-2026-10-23-016`
- Type: `ticker`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-016/quantum-simulation-finds-a-cheaper-contraction-order

**Dek:** ArcTN's Light preset cut planned FLOPs 61.32 percent against one compared baseline while planning 1.62 times faster.

ArcTN combines randomized-greedy construction, subtree reconfiguration and parallel tempering, then slices tensors to fit memory limits. Its relative gains vary by network and comparator, but the reported suite shows lower planning-plus-execution cost on the common completed runs.

### Why it matters {#why-it-matters-mp-2026-10-23-016}

ArcTN's Light preset cut planned FLOPs 61.32 percent against one compared baseline while planning 1.62 times faster.

### Limits and context {#limitations-mp-2026-10-23-016}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-10-23-016}

- ArcTN's Light preset cut planned FLOPs 61.32 percent against one compared baseline while planning 1.62 times faster. [source-2026-10-23-018]

## 19. A $35 Robot Skin Feels Sideways Force {#mp-2026-10-23-017}

- Story ID: `mp-2026-10-23-017`
- Type: `ticker`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-017/a-35-robot-skin-feels-sideways-force

**Dek:** VecTile is six millimeters thin, flexible and built in-house; silicone nubs translate three-axis force into distinguishable pressure patterns.

The 4-by-4 array reported 0.39-newton normal-force and 0.16-newton shear-force RMSE in controlled tests. It stayed calibrated while curved, stopped a scalpel before it punctured a balloon, and improved a plug-insertion policy. Designs are promised after paper acceptance.

### Why it matters {#why-it-matters-mp-2026-10-23-017}

VecTile is six millimeters thin, flexible and built in-house; silicone nubs translate three-axis force into distinguishable pressure patterns.

### Limits and context {#limitations-mp-2026-10-23-017}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-10-23-017}

- VecTile is six millimeters thin, flexible and built in-house; silicone nubs translate three-axis force into distinguishable pressure patterns. [source-2026-10-23-019]

## 20. The Speech Evaluator Was Mishearing the Rare Words {#mp-2026-10-23-018}

- Story ID: `mp-2026-10-23-018`
- Type: `ticker`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-018/the-speech-evaluator-was-mishearing-the-rare-words

**Dek:** Evaluation recognizers missed 33 to 67 percent of very rare words even in human recordings used as the control.

Adding two recognizer families cut apparent very-rare-word TTS error from roughly 10 to 13 percent down to 3 to 5 percent on words at least one system heard. Yet masking recognizer misses introduced its own bias, and the authors conclude that listening tests remain necessary.

### Why it matters {#why-it-matters-mp-2026-10-23-018}

Evaluation recognizers missed 33 to 67 percent of very rare words even in human recordings used as the control.

### Limits and context {#limitations-mp-2026-10-23-018}

- Yet masking recognizer misses introduced its own bias, and the authors conclude that listening tests remain necessary.

### Claims and sources {#claims-mp-2026-10-23-018}

- Evaluation recognizers missed 33 to 67 percent of very rare words even in human recordings used as the control. [source-2026-10-23-020] — Qualification: Yet masking recognizer misses introduced its own bias, and the authors conclude that listening tests remain necessary.

## 21. The Causal Mask Passed While Future Tokens Leaked Elsewhere {#mp-2026-10-23-019}

- Story ID: `mp-2026-10-23-019`
- Type: `ticker`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-019/the-causal-mask-passed-while-future-tokens-leaked-elsewhere

**Dek:** A full-sequence routing summary made validation loss look dramatically better even as free-running generation deteriorated.

Prefix-invariance checks localized the leak to a dynamic routing-query path outside causal attention. Two repaired models lost the apparent advantage and returned to the control range, showing why teacher-forced loss needs whole-model causal and generation audits.

### Why it matters {#why-it-matters-mp-2026-10-23-019}

A full-sequence routing summary made validation loss look dramatically better even as free-running generation deteriorated.

### Limits and context {#limitations-mp-2026-10-23-019}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-10-23-019}

- A full-sequence routing summary made validation loss look dramatically better even as free-running generation deteriorated. [source-2026-10-23-021]

## 22. Space Drums 2.0 {#mp-2026-10-23-020}

- Story ID: `mp-2026-10-23-020`
- Type: `invention_desk`
- Classification: `editorial`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-020/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-2026-10-23-020}

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

### Limits and context {#limitations-mp-2026-10-23-020}

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

### Claims and sources {#claims-mp-2026-10-23-020}

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

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

- Story ID: `mp-2026-10-23-021`
- Type: `invention_desk`
- Classification: `editorial`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-021/3ric

**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-2026-10-23-021}

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

### Limits and context {#limitations-mp-2026-10-23-021}

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

### Claims and sources {#claims-mp-2026-10-23-021}

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

## 24. M3-CRETE {#mp-2026-10-23-022}

- Story ID: `mp-2026-10-23-022`
- Type: `invention_desk`
- Classification: `editorial`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-022/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-2026-10-23-022}

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

### Limits and context {#limitations-mp-2026-10-23-022}

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

### Claims and sources {#claims-mp-2026-10-23-022}

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

## 25. DeskNode {#mp-2026-10-23-023}

- Story ID: `mp-2026-10-23-023`
- Type: `invention_desk`
- Classification: `editorial`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-023/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-2026-10-23-023}

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

### Limits and context {#limitations-mp-2026-10-23-023}

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

### Claims and sources {#claims-mp-2026-10-23-023}

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

## 26. The First Paid Slot {#mp-2026-10-23-024}

- Story ID: `mp-2026-10-23-024`
- Type: `invention_desk`
- Classification: `house_example`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-024/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-2026-10-23-024}

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

### Limits and context {#limitations-mp-2026-10-23-024}

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

### Claims and sources {#claims-mp-2026-10-23-024}

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

## 27. Put Your Project on the Desk {#mp-2026-10-23-025}

- Story ID: `mp-2026-10-23-025`
- Type: `invention_desk`
- Classification: `house_example`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-10-23-025/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-2026-10-23-025}

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

### Limits and context {#limitations-mp-2026-10-23-025}

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

### Claims and sources {#claims-mp-2026-10-23-025}

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

## Normalized sources

- **source-2026-10-23-001:** [arXiv preprint 2610.31651, original v1](https://arxiv.org/abs/2610.31651v1) — arXiv; primary_research
- **source-2026-10-23-002:** [arXiv preprint 2610.31662, original v1](https://arxiv.org/abs/2610.31662v1) — arXiv; primary_research
- **source-2026-10-23-003:** [arXiv preprint 2610.31661, original v1](https://arxiv.org/abs/2610.31661v1) — arXiv; primary_research
- **source-2026-10-23-004:** [arXiv preprint 2610.31659, original v1](https://arxiv.org/abs/2610.31659v1) — arXiv; primary_research
- **source-2026-10-23-005:** [arXiv preprint 2610.31650, original v1](https://arxiv.org/abs/2610.31650v1) — arXiv; primary_research
- **source-2026-10-23-006:** [arXiv preprint 2610.31667, original v1](https://arxiv.org/abs/2610.31667v1) — arXiv; primary_research
- **source-2026-10-23-007:** [arXiv preprint 2610.31670, original v1](https://arxiv.org/abs/2610.31670v1) — arXiv; primary_research
- **source-2026-10-23-008:** [arXiv preprint 2610.31636, original v1](https://arxiv.org/abs/2610.31636v1) — arXiv; primary_research
- **source-2026-10-23-009:** [arXiv preprint 2610.31657, original v1](https://arxiv.org/abs/2610.31657v1) — arXiv; primary_research
- **source-2026-10-23-010:** [arXiv preprint 2610.31639, original v1](https://arxiv.org/abs/2610.31639v1) — arXiv; primary_research
- **source-2026-10-23-011:** [arXiv preprint 2610.31624, original v1](https://arxiv.org/abs/2610.31624v1) — arXiv; primary_research
- **source-2026-10-23-012:** [arXiv preprint 2610.31616, original v1](https://arxiv.org/abs/2610.31616v1) — arXiv; primary_research
- **source-2026-10-23-013:** [arXiv preprint 2610.31588, original v1](https://arxiv.org/abs/2610.31588v1) — arXiv; primary_research
- **source-2026-10-23-014:** [arXiv preprint 2610.31613, original v1](https://arxiv.org/abs/2610.31613v1) — arXiv; primary_research
- **source-2026-10-23-015:** [arXiv preprint 2610.31593, original v1](https://arxiv.org/abs/2610.31593v1) — arXiv; primary_research
- **source-2026-10-23-016:** [arXiv preprint 2610.31579, original v1](https://arxiv.org/abs/2610.31579v1) — arXiv; primary_research
- **source-2026-10-23-017:** [arXiv preprint 2610.31632, original v1](https://arxiv.org/abs/2610.31632v1) — arXiv; primary_research
- **source-2026-10-23-018:** [arXiv preprint 2610.31554, original v1](https://arxiv.org/abs/2610.31554v1) — arXiv; primary_research
- **source-2026-10-23-019:** [arXiv preprint 2610.31472, original v1](https://arxiv.org/abs/2610.31472v1) — arXiv; primary_research
- **source-2026-10-23-020:** [arXiv preprint 2610.31470, original v1](https://arxiv.org/abs/2610.31470v1) — arXiv; primary_research
- **source-2026-10-23-021:** [arXiv preprint 2610.31522, original v1](https://arxiv.org/abs/2610.31522v1) — arXiv; primary_research