---
schema_version: "1.0.0"
edition_id: "mp-2026-09-16-morning-0069"
published_at: "2026-09-16T09:00:00.000-04:00"
modified_at: "2026-09-16T09:00:00.000-04:00"
canonical_url: "https://themachinepress.com/edition/2026-09-16"
story_count: 27
lead_story_id: "mp-2026-09-16-001"
---

# The Machine Press — Morning edition

Edition ID: `mp-2026-09-16-morning-0069`  
Published: 2026-09-16T09:00:00.000-04:00  
Canonical edition: https://themachinepress.com/edition/2026-09-16

Across 200 FreeCAD tasks, the strongest tested computer-use agent passed 17.5%; expert reference work passed 87.0%.

## 1. The Best CAD Agent Finished One Task in Six {#mp-2026-09-16-001}

- Story ID: `mp-2026-09-16-001`
- Type: `lead`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-001/the-best-cad-agent-finished-one-task-in-six

**Dek:** Across 200 FreeCAD tasks, the strongest tested computer-use agent passed 17.5%; expert reference work passed 87.0%.

Computer-use benchmarks often stop when an interface appears to accept the right clicks. CADWorld checks the persistent engineering artifact instead. Its 200 FreeCAD tasks span 11 workflows, from sketches and assemblies to CAM, finite-element analysis, meshes and technical drawings. Executable checks inspect geometry, dimensions, parametric structure, constraints and downstream manufacturing state. Across seven current agents, the strongest passed 17.5% of the full suite versus 87.0% for an expert reference. Stronger agents increasingly reached plausible-looking outputs but failed structural and construction-process requirements, exposing a gap between GUI fluency and reliable engineering work.

### Why it matters {#why-it-matters-mp-2026-09-16-001}

Across 200 FreeCAD tasks, the strongest tested computer-use agent passed 17.5%; expert reference work passed 87.0%.

### Limits and context {#limitations-mp-2026-09-16-001}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-09-16-001}

- Across 200 FreeCAD tasks, the strongest tested computer-use agent passed 17.5%; expert reference work passed 87.0%. [source-2026-09-16-001]

## 2. One Hundred Qubits Settled a Dissipative Chain {#mp-2026-09-16-002}

- Story ID: `mp-2026-09-16-002`
- Type: `secondary`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-002/one-hundred-qubits-settled-a-dissipative-chain

**Dek:** A superconducting processor simulated 50 sites at depths up to 1,700 entangling gates and mapped 117 hardware points.

Open quantum systems can settle into steady states with no equilibrium counterpart, but their density matrices grow too quickly for controlled large-scale calculations. The team mapped a dissipative spin-1/2 Heisenberg chain onto 100 simultaneously active qubits on IBM's Kingston processor, using Stinespring dilation to represent 50 sites at entangling-gate depths up to 1,700. From 117 hardware measurements, they resolved ferromagnetic, antiferromagnetic, spin-density-wave and paramagnetic regions. Because the modeled dissipation continually erases some errors, hardware noise acts as a weaker competing bath. The result is a hardware study of one benchmark model, not a general claim of quantum advantage.

### Why it matters {#why-it-matters-mp-2026-09-16-002}

A superconducting processor simulated 50 sites at depths up to 1,700 entangling gates and mapped 117 hardware points.

### Limits and context {#limitations-mp-2026-09-16-002}

- The result is a hardware study of one benchmark model, not a general claim of quantum advantage.

### Claims and sources {#claims-mp-2026-09-16-002}

- A superconducting processor simulated 50 sites at depths up to 1,700 entangling gates and mapped 117 hardware points. [source-2026-09-16-002] — Qualification: The result is a hardware study of one benchmark model, not a general claim of quantum advantage.

## 3. Pruning Followed the Model's Geometry {#mp-2026-09-16-003}

- Story ID: `mp-2026-09-16-003`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-003/pruning-followed-the-model-s-geometry

**Dek:** A Fisher-geodesic hierarchy beat magnitude and local-Fisher pruning across every tested architecture and dataset combination.

Setting a neural-network parameter to zero moves the model onto a constrained surface. This paper measures that displacement with the Fisher information geometry, then derives a hierarchy from simple magnitude pruning toward progressively more faithful geodesic approximations. Across fully connected networks and vision transformers on MNIST and CIFAR-10, over the complete zero-to-100% pruning range and five seeds, the proposed schemes outperformed magnitude and local-Fisher pruning on both accuracy and Matthews correlation. Intermediate approximations retained much of the gain with lower computational cost; the evidence remains limited to the reported models and datasets.

### Why it matters {#why-it-matters-mp-2026-09-16-003}

A Fisher-geodesic hierarchy beat magnitude and local-Fisher pruning across every tested architecture and dataset combination.

### Limits and context {#limitations-mp-2026-09-16-003}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-09-16-003}

- A Fisher-geodesic hierarchy beat magnitude and local-Fisher pruning across every tested architecture and dataset combination. [source-2026-09-16-003]

## 4. A Small Corrector Fixed Errors Without Rewriting the Base Model {#mp-2026-09-16-004}

- Story ID: `mp-2026-09-16-004`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-004/a-small-corrector-fixed-errors-without-rewriting-the-base-model

**Dek:** A 34-million-parameter module corrected 53.3% of tested errors while the matched evaluations showed no base-capability loss.

CRN v2 leaves a 4.65-billion-parameter language model frozen and learns a roughly 34-million-parameter logit correction layer on 83,400 pairs. On a 60-question domain exam it corrected 53.3% of base-model errors; a reworded version reached 43.3%. A matched LoRA baseline corrected more errors but lost 30% to 75% on the same small capability checks. Lowering the preservation penalty sharply reduced correction. The experiment supports frozen-base adjustment as a tradeoff, but its 60-question domain exam and small benchmark samples are not evidence of broad capability preservation.

### Why it matters {#why-it-matters-mp-2026-09-16-004}

A 34-million-parameter module corrected 53.3% of tested errors while the matched evaluations showed no base-capability loss.

### Limits and context {#limitations-mp-2026-09-16-004}

- The experiment supports frozen-base adjustment as a tradeoff, but its 60-question domain exam and small benchmark samples are not evidence of broad capability preservation.

### Claims and sources {#claims-mp-2026-09-16-004}

- A 34-million-parameter module corrected 53.3% of tested errors while the matched evaluations showed no base-capability loss. [source-2026-09-16-004] — Qualification: The experiment supports frozen-base adjustment as a tradeoff, but its 60-question domain exam and small benchmark samples are not evidence of broad capability preservation.

## 5. Six GPUs Matched Seven After the Router Was Calibrated {#mp-2026-09-16-005}

- Story ID: `mp-2026-09-16-005`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-005/six-gpus-matched-seven-after-the-router-was-calibrated

**Dek:** A learned request router reached 0.864 mean goodput and matched round robin's seven-GPU result with six GPUs.

Disaggregated serving separates prompt prefill from token decoding, making request placement a latency and capacity problem. This study predicts completion time from prompt length, expected output, post-admission cache pressure and service class, then calibrates the scorer on eight NVIDIA A40 GPUs. Across three bursty traces, mean goodput reached 0.864 versus 0.835 to 0.847 for round robin, least-loaded and length heuristics. Calibration mattered: simulator-only constants erased 4.5 goodput points and much of the tail-latency gain. Under extreme scarcity, greedy routing concentrated work too aggressively, marking a clear operating boundary.

### Why it matters {#why-it-matters-mp-2026-09-16-005}

A learned request router reached 0.864 mean goodput and matched round robin's seven-GPU result with six GPUs.

### Limits and context {#limitations-mp-2026-09-16-005}

- Calibration mattered: simulator-only constants erased 4.5 goodput points and much of the tail-latency gain.

### Claims and sources {#claims-mp-2026-09-16-005}

- A learned request router reached 0.864 mean goodput and matched round robin's seven-GPU result with six GPUs. [source-2026-09-16-005] — Qualification: Calibration mattered: simulator-only constants erased 4.5 goodput points and much of the tail-latency gain.

## 6. Tier Capacity Beat Clever Cache Prediction {#mp-2026-09-16-006}

- Story ID: `mp-2026-09-16-006`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-006/tier-capacity-beat-clever-cache-prediction

**Dek:** GPU, CPU and SSD tiering supported 73.02 times more sessions per GPU; predicted reuse and prefetching did not earn their cost.

Long-lived chats and agent loops fill expensive GPU memory with key-value cache blocks. A calibrated simulator compared recency, reuse frequency, predicted reuse and look-ahead prefetch across GPU HBM, CPU DRAM and SSD. The three-tier capacity model supported 73.02 times more concurrent sessions per GPU and cut cost per session 62.04 times, but those gains came from capacity rather than sophisticated placement. Recency minimized chat migration while reuse frequency led for agents and document work. Even an oracle prefetch policy failed to beat no prefetch on migration traffic, challenging recommendations that prediction alone improves this hierarchy.

### Why it matters {#why-it-matters-mp-2026-09-16-006}

GPU, CPU and SSD tiering supported 73.02 times more sessions per GPU; predicted reuse and prefetching did not earn their cost.

### Limits and context {#limitations-mp-2026-09-16-006}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-09-16-006}

- GPU, CPU and SSD tiering supported 73.02 times more sessions per GPU; predicted reuse and prefetching did not earn their cost. [source-2026-09-16-006]

## 7. A Frozen Model Changed Its Scientific Working Mode {#mp-2026-09-16-007}

- Story ID: `mp-2026-09-16-007`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-007/a-frozen-model-changed-its-scientific-working-mode

**Dek:** A controller switched a frozen model among exploration, execution and reassessment using low-dimensional internal interventions.

Metacognitive Steering looks for process-level signals in scientist interaction traces rather than training only on finished scientific outputs. The authors identify a coordinated control surface across middle layers of a frozen mixture-of-experts model, then read its current regime and compose interventions for exploration, procedural convergence or critical reassessment. They report more sustained exploration, explicit pruning and evidence-responsive synthesis, and describe an autonomous system that reproduced eight BlueZ vulnerabilities and directed a rocket engineering project. Those demonstrations are author-reported case studies; independent replication is needed before treating the control method as reliable scientific judgment.

### Why it matters {#why-it-matters-mp-2026-09-16-007}

A controller switched a frozen model among exploration, execution and reassessment using low-dimensional internal interventions.

### Limits and context {#limitations-mp-2026-09-16-007}

- Metacognitive Steering looks for process-level signals in scientist interaction traces rather than training only on finished scientific outputs.

### Claims and sources {#claims-mp-2026-09-16-007}

- A controller switched a frozen model among exploration, execution and reassessment using low-dimensional internal interventions. [source-2026-09-16-007] — Qualification: Metacognitive Steering looks for process-level signals in scientist interaction traces rather than training only on finished scientific outputs.

## 8. Zeros Pushed Ternary Weights Below 1.58 Bits {#mp-2026-09-16-008}

- Story ID: `mp-2026-09-16-008`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-008/zeros-pushed-ternary-weights-below-1-58-bits

**Dek:** BITCOS reached 1.485 bits per weight on the sparsest tested model and improved decode throughput up to 1.27 times.

Ternary models are usually packed as if minus one, zero and plus one were equally likely. Measurements across 29 models found zeros reaching 51.5% of weights. BITCOS stores a presence bitmap plus compacted signs, costing two minus the zero density in bits per element. It beat five-trit packing in 26 of 29 tested models and reached 1.485 bits per weight on the sparsest. Optimized unpacking improved production kernel throughput up to 1.28 times and end-to-end decode up to 1.18 times on CPUs and 1.27 times on GPUs in the reported tests.

### Why it matters {#why-it-matters-mp-2026-09-16-008}

BITCOS reached 1.485 bits per weight on the sparsest tested model and improved decode throughput up to 1.27 times.

### Limits and context {#limitations-mp-2026-09-16-008}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-09-16-008}

- BITCOS reached 1.485 bits per weight on the sparsest tested model and improved decode throughput up to 1.27 times. [source-2026-09-16-008]

## 9. The Answer Key Recomputed Itself From Live Data {#mp-2026-09-16-009}

- Story ID: `mp-2026-09-16-009`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-009/the-answer-key-recomputed-itself-from-live-data

**Dek:** Executable ground truth improved expert-judge agreement 29% by MCC while reducing tokens 16% per test case.

Static answers go stale when a data-science agent is asked about live audience or business data. This framework encodes the expected result as a function that recomputes directly from the underlying database at evaluation time, then scores atomic facts regardless of whether the agent replies in prose, a table or HTML. In an author-reported human agreement study using a synthetic database shaped like a production system, ground-truth-as-code improved Matthews correlation with expert labels 29% over natural-language references and used 16% fewer tokens. A self-directed baseline without explicit ground truth was anti-correlated with human judgment.

### Why it matters {#why-it-matters-mp-2026-09-16-009}

Executable ground truth improved expert-judge agreement 29% by MCC while reducing tokens 16% per test case.

### Limits and context {#limitations-mp-2026-09-16-009}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-09-16-009}

- Executable ground truth improved expert-judge agreement 29% by MCC while reducing tokens 16% per test case. [source-2026-09-16-009]

## 10. The Surgical Robot Tracked the Phantom After Every Cut {#mp-2026-09-16-010}

- Story ID: `mp-2026-09-16-010`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-010/the-surgical-robot-tracked-the-phantom-after-every-cut

**Dek:** Eight autonomous phantom resections produced negative margins across 77 cuts, with 1.61-millimeter mean absolute margin error.

Soft-tissue surgery changes the anatomy the robot is trying to track. This supervised research platform infers complete tumor, margin and kidney geometry from a single partial point cloud, updating its occupancy model as hydrogel tissue deforms and is cut. In eight consecutive open partial-nephrectomy phantom trials, dual robotic arms completed 77 electrosurgical cuts; every cut achieved a negative margin, with 1.61 ± 0.48 millimeters mean absolute margin error. These were patient-derived phantoms in a controlled open setting, not operations on people, and the results do not establish clinical autonomy.

### Why it matters {#why-it-matters-mp-2026-09-16-010}

Eight autonomous phantom resections produced negative margins across 77 cuts, with 1.61-millimeter mean absolute margin error.

### Limits and context {#limitations-mp-2026-09-16-010}

- These were patient-derived phantoms in a controlled open setting, not operations on people, and the results do not establish clinical autonomy.

### Claims and sources {#claims-mp-2026-09-16-010}

- Eight autonomous phantom resections produced negative margins across 77 cuts, with 1.61-millimeter mean absolute margin error. [source-2026-09-16-010] — Qualification: These were patient-derived phantoms in a controlled open setting, not operations on people, and the results do not establish clinical autonomy.

## 11. The Robot Learned 'Gently' as an Execution Control {#mp-2026-09-16-011}

- Story ID: `mp-2026-09-16-011`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-011/the-robot-learned-gently-as-an-execution-control

**Dek:** Modifier-conditioned decoding improved force-direction following on a real whiteboard-wiping task while retaining speed control.

Contact-rich imitation learning usually reproduces an action without giving the operator a direct way to ask for slower, faster, gentler or firmer execution. Bi-MoDe injects a constrained modifier latent into every layer of a Transformer action decoder, allowing directives to alter action chunks. On a physical whiteboard-wiping task with combinations of temporal and force modifiers, it improved physical-directive following over the action-chunking baseline while maintaining comparable temporal control. The experiment demonstrates one task and robot setup, not a general natural-language safety interface.

### Why it matters {#why-it-matters-mp-2026-09-16-011}

Modifier-conditioned decoding improved force-direction following on a real whiteboard-wiping task while retaining speed control.

### Limits and context {#limitations-mp-2026-09-16-011}

- The experiment demonstrates one task and robot setup, not a general natural-language safety interface.

### Claims and sources {#claims-mp-2026-09-16-011}

- Modifier-conditioned decoding improved force-direction following on a real whiteboard-wiping task while retaining speed control. [source-2026-09-16-011] — Qualification: The experiment demonstrates one task and robot setup, not a general natural-language safety interface.

## 12. Torque Feedback Moved Onto the Underwater Gripper {#mp-2026-09-16-012}

- Story ID: `mp-2026-09-16-012`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-012/torque-feedback-moved-onto-the-underwater-gripper

**Dek:** Twenty operators reported less attention-shifting burden with registered mixed-reality overlays and similar torque regulation.

Underwater bilateral teleoperation can show reaction torque on a monitor, forcing the operator to shift attention between feedback and the workspace. MR-GLi spatially registers the torque indicator and wrist-camera view to the robot gripper. Twenty participants lifted and moved rigid and compliant objects in a counterbalanced comparison against the same feedback on a two-dimensional monitor. Torque-regulation performance remained similar, while subjective reports indicated less burden from shifting attention. The result supports information placement, not superior task performance or field readiness.

### Why it matters {#why-it-matters-mp-2026-09-16-012}

Twenty operators reported less attention-shifting burden with registered mixed-reality overlays and similar torque regulation.

### Limits and context {#limitations-mp-2026-09-16-012}

- The result supports information placement, not superior task performance or field readiness.

### Claims and sources {#claims-mp-2026-09-16-012}

- Twenty operators reported less attention-shifting burden with registered mixed-reality overlays and similar torque regulation. [source-2026-09-16-012] — Qualification: The result supports information placement, not superior task performance or field readiness.

## 13. Factory Robots Formed Temporary Coalitions With Local Information {#mp-2026-09-16-013}

- Story ID: `mp-2026-09-16-013`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-013/factory-robots-formed-temporary-coalitions-with-local-information

**Dek:** AssemblyGrid separates task success from learning reward across flow, coalition and concurrency workloads.

Flexible production can require robots to route material, share resources, form temporary teams and work concurrently while each sees only local information. AssemblyGrid v1 combines those constraints with geometry-dependent feasibility in one repeatable benchmark. It defines three workload families at three scenario levels, with executable conformance checks and success metrics independent of any learning reward. Centralized references, structured decentralized controllers and IPPO, MAPPO and QMIX experiments all produced productive behavior. The benchmark is a controlled task-level environment, not evidence of autonomous deployment on a factory floor.

### Why it matters {#why-it-matters-mp-2026-09-16-013}

AssemblyGrid separates task success from learning reward across flow, coalition and concurrency workloads.

### Limits and context {#limitations-mp-2026-09-16-013}

- Flexible production can require robots to route material, share resources, form temporary teams and work concurrently while each sees only local information.
- The benchmark is a controlled task-level environment, not evidence of autonomous deployment on a factory floor.

### Claims and sources {#claims-mp-2026-09-16-013}

- AssemblyGrid separates task success from learning reward across flow, coalition and concurrency workloads. [source-2026-09-16-013] — Qualification: Flexible production can require robots to route material, share resources, form temporary teams and work concurrently while each sees only local information.

## 14. The Manipulation Skill Declared Its Geometry First {#mp-2026-09-16-014}

- Story ID: `mp-2026-09-16-014`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-014/the-manipulation-skill-declared-its-geometry-first

**Dek:** Without new demonstrations, geometric contracts reached 88.24% pick-and-place and 75.00% average functional-task success.

ManiSkillFormer replaces an end-to-end visuomotor policy with task-conditioned geometric contracts. Each skill declares the object keypoints, surface normals and other primitives perception must find; language-model agents generate contracts and motion templates inside human-defined skill structures. On a dual-arm robot, the authors report 88.24% zero-shot pick-and-place across eight object categories and 30 instances, 75.00% average success on unscrewing, pouring, pressing and folding, and 50% to 80% completion on three long-horizon tasks. The results are confined to the reported platform and skill library.

### Why it matters {#why-it-matters-mp-2026-09-16-014}

Without new demonstrations, geometric contracts reached 88.24% pick-and-place and 75.00% average functional-task success.

### Limits and context {#limitations-mp-2026-09-16-014}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-09-16-014}

- Without new demonstrations, geometric contracts reached 88.24% pick-and-place and 75.00% average functional-task success. [source-2026-09-16-014]

## 15. The Route Planner Reconsidered Only Its Uncertain Waypoints {#mp-2026-09-16-026}

- Story ID: `mp-2026-09-16-026`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-026/the-route-planner-reconsidered-only-its-uncertain-waypoints

**Dek:** Stochastic consensus plus selective reconsideration cut mean displacement error 25.1% across 400 held-out aerial queries.

A vision-language waypoint planner normally emits one route without saying how reliable it is. UDAV samples multiple trajectories from aerial imagery, chooses the medoid and treats their spatial spread as uncertainty. When an interior waypoint crosses a threshold, it invokes a reconsideration stage. Across 400 held-out queries from two UAV flights, medoid selection cut mean average displacement error from 147.4 to 115.9 pixels; the complete planner reached 110.4 pixels, a 25.1% reduction, while returning valid trajectories for all queries. Pixel error on two flights is not evidence of safe autonomous navigation.

### Why it matters {#why-it-matters-mp-2026-09-16-026}

Stochastic consensus plus selective reconsideration cut mean displacement error 25.1% across 400 held-out aerial queries.

### Limits and context {#limitations-mp-2026-09-16-026}

- Pixel error on two flights is not evidence of safe autonomous navigation.

### Claims and sources {#claims-mp-2026-09-16-026}

- Stochastic consensus plus selective reconsideration cut mean displacement error 25.1% across 400 held-out aerial queries. [source-2026-09-16-015] — Qualification: Pixel error on two flights is not evidence of safe autonomous navigation.

## 16. A Hot Jupiter's Clouds Vanished Where Its Dayside Burned Hottest {#mp-2026-09-16-027}

- Story ID: `mp-2026-09-16-027`
- Type: `dispatch`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-027/a-hot-jupiter-s-clouds-vanished-where-its-dayside-burned-hottest

**Dek:** JWST separated reflected light from heat and measured transport becoming 7.1 ± 1.9 degrees more efficient per pressure decade.

A phase curve follows a planet through its orbit, mixing reflected starlight with thermal emission. JWST's NIRSpec PRISM separated those components for NGTS-10 Ab from 0.5 to 5.5 micrometers, producing joint temperature and reflectance maps. The two were anti-correlated: models best explained the pattern with micron-scale silicate clouds evaporating across the hottest eastern substellar region and weak atmospheric drag. The thermal offset changed by 7.1 ± 1.9 degrees per pressure decade, indicating stronger heat transport deeper in the atmosphere. The interpretation depends on retrieval and circulation models rather than direct images of clouds.

### Why it matters {#why-it-matters-mp-2026-09-16-027}

JWST separated reflected light from heat and measured transport becoming 7.1 ± 1.9 degrees more efficient per pressure decade.

### Limits and context {#limitations-mp-2026-09-16-027}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-09-16-027}

- JWST separated reflected light from heat and measured transport becoming 7.1 ± 1.9 degrees more efficient per pressure decade. [source-2026-09-16-016]

## 17. Continuous-Variable QKD Closed Its Coherent-Attack Gap {#mp-2026-09-16-015}

- Story ID: `mp-2026-09-16-015`
- Type: `ticker`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-015/continuous-variable-qkd-closed-its-coherent-attack-gap

**Dek:** A composable finite-size proof retained positive modeled key rates beyond 70 kilometers with imperfect detectors.

Discrete-modulated continuous-variable quantum key distribution has used experimentally attractive telecom components without a complete finite-size proof against coherent attacks. This work introduces an infinite-dimensional entropy accumulation theorem and a dimension-reduction method, covering imperfect detectors and fixed- or variable-length protocols. The computed bounds recover known asymptotic rates and remain positive beyond 70 kilometers for the reported block sizes and parameters. It is a security proof and numerical result, not a field deployment.

### Why it matters {#why-it-matters-mp-2026-09-16-015}

A composable finite-size proof retained positive modeled key rates beyond 70 kilometers with imperfect detectors.

### Limits and context {#limitations-mp-2026-09-16-015}

- The computed bounds recover known asymptotic rates and remain positive beyond 70 kilometers for the reported block sizes and parameters.
- It is a security proof and numerical result, not a field deployment.

### Claims and sources {#claims-mp-2026-09-16-015}

- A composable finite-size proof retained positive modeled key rates beyond 70 kilometers with imperfect detectors. [source-2026-09-16-017] — Qualification: The computed bounds recover known asymptotic rates and remain positive beyond 70 kilometers for the reported block sizes and parameters.

## 18. A Diamond Spin Ensemble Revealed Its Own Projection Noise {#mp-2026-09-16-016}

- Story ID: `mp-2026-09-16-016`
- Type: `ticker`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-016/a-diamond-spin-ensemble-revealed-its-own-projection-noise

**Dek:** Nuclear-memory readout improved fidelity nearly tenfold and resolved interaction-driven anisotropy.

A two-dimensional nitrogen-vacancy ensemble in diamond was read repeatedly through intrinsic nitrogen-15 nuclear memory at about 0.3 tesla. The method improved readout fidelity by nearly an order of magnitude, enough to resolve coherent-state quantum projection noise and watch dipolar interactions shear it into an anisotropic profile. The experiment opens a route toward directly measuring spin squeezing and entanglement-enhanced sensing, but does not report those later sensing gains yet.

### Why it matters {#why-it-matters-mp-2026-09-16-016}

Nuclear-memory readout improved fidelity nearly tenfold and resolved interaction-driven anisotropy.

### Limits and context {#limitations-mp-2026-09-16-016}

- The experiment opens a route toward directly measuring spin squeezing and entanglement-enhanced sensing, but does not report those later sensing gains yet.

### Claims and sources {#claims-mp-2026-09-16-016}

- Nuclear-memory readout improved fidelity nearly tenfold and resolved interaction-driven anisotropy. [source-2026-09-16-018] — Qualification: The experiment opens a route toward directly measuring spin squeezing and entanglement-enhanced sensing, but does not report those later sensing gains yet.

## 19. Hyperbolic Color Codes Reached Constant Rate and Polynomial Distance {#mp-2026-09-16-017}

- Story ID: `mp-2026-09-16-017`
- Type: `ticker`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-017/hyperbolic-color-codes-reached-constant-rate-and-polynomial-distance

**Dek:** The construction combines high-rate qLDPC scaling with color-code structure in dimensions four and above.

Earlier hyperbolic color-code constructions reached at most logarithmic code distance. Building on arithmetic hyperbolic manifolds, this work constructs families with constant encoding rate and polynomial distance in even dimensions, with polynomial logical-qubit and distance scaling in the other covered cases. The paper derives explicit lower-bound exponents by dimension and code type. This is a mathematical construction and testbed for fault-tolerant protocols, not a hardware demonstration.

### Why it matters {#why-it-matters-mp-2026-09-16-017}

The construction combines high-rate qLDPC scaling with color-code structure in dimensions four and above.

### Limits and context {#limitations-mp-2026-09-16-017}

- This is a mathematical construction and testbed for fault-tolerant protocols, not a hardware demonstration.

### Claims and sources {#claims-mp-2026-09-16-017}

- The construction combines high-rate qLDPC scaling with color-code structure in dimensions four and above. [source-2026-09-16-019] — Qualification: This is a mathematical construction and testbed for fault-tolerant protocols, not a hardware demonstration.

## 20. Eight Galaxies Kept Helium but Hid Their Metals {#mp-2026-09-16-018}

- Story ID: `mp-2026-09-16-018`
- Type: `ticker`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-018/eight-galaxies-kept-helium-but-hid-their-metals

**Dek:** A search of 109,545 emitters found eight tentative Population III-like candidates near redshift two.

Researchers searched 109,545 high-confidence HETDEX Lyman-alpha emitters for strong narrow helium emission without detected ultraviolet metal lines. Eight candidates produced a stacked helium equivalent width of 28.7 ± 6.7 angstroms and no significant nitrogen, carbon or oxygen features at the reported limits. The combination is difficult to reproduce with several enriched-star, active-galaxy and shock alternatives. The authors call them very metal-poor or Population III-like candidates; deeper individual ultraviolet and optical spectra are required for confirmation.

### Why it matters {#why-it-matters-mp-2026-09-16-018}

A search of 109,545 emitters found eight tentative Population III-like candidates near redshift two.

### Limits and context {#limitations-mp-2026-09-16-018}

- No additional limitation was separately recorded.

### Claims and sources {#claims-mp-2026-09-16-018}

- A search of 109,545 emitters found eight tentative Population III-like candidates near redshift two. [source-2026-09-16-020]

## 21. A Messier Microwave Sky Barely Moved the Cosmology {#mp-2026-09-16-019}

- Story ID: `mp-2026-09-16-019`
- Type: `ticker`
- Classification: `editorial`
- Content status: `new`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-019/a-messier-microwave-sky-barely-moved-the-cosmology

**Dek:** Residual foreground power rose 50% over idealized estimates, while 11-parameter uncertainties increased less than 7%.

CMB-HD foreground simulations cover 100 square degrees at 2.5-arcsecond resolution and include infrared sources, radio galaxies and galaxy clusters. After source detection and cleaning, residual foreground-plus-noise power at 90 and 148 gigahertz was 50% higher than earlier idealized assumptions. Even so, forecast uncertainties across an 11-parameter cosmology grew by less than 7%, helped by polarization, cross spectra and lensing. The released simulations forecast a proposed survey; they are not observed CMB-HD data.

### Why it matters {#why-it-matters-mp-2026-09-16-019}

Residual foreground power rose 50% over idealized estimates, while 11-parameter uncertainties increased less than 7%.

### Limits and context {#limitations-mp-2026-09-16-019}

- The released simulations forecast a proposed survey; they are not observed CMB-HD data.

### Claims and sources {#claims-mp-2026-09-16-019}

- Residual foreground power rose 50% over idealized estimates, while 11-parameter uncertainties increased less than 7%. [source-2026-09-16-021] — Qualification: The released simulations forecast a proposed survey; they are not observed CMB-HD data.

## 22. SATURNIX {#mp-2026-09-16-020}

- Story ID: `mp-2026-09-16-020`
- Type: `invention_desk`
- Classification: `editorial`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-020/saturnix

**Dek:** Builds a tactile digital camera around a Raspberry Pi Zero 2 W, an autofocus sensor, a small viewfinder, and mechanical-switch controls while publishing the software and printable hardware files.

Builds a tactile digital camera around a Raspberry Pi Zero 2 W, an autofocus sensor, a small viewfinder, and mechanical-switch controls while publishing the software and printable hardware files.

### Why it matters {#why-it-matters-mp-2026-09-16-020}

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

### Limits and context {#limitations-mp-2026-09-16-020}

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

### Claims and sources {#claims-mp-2026-09-16-020}

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

## 23. WURB-2026 {#mp-2026-09-16-021}

- Story ID: `mp-2026-09-16-021`
- Type: `invention_desk`
- Classification: `editorial`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-021/wurb-2026

**Dek:** Combines a Raspberry Pi-class computer, an ultrasonic microphone, local storage, and a web interface into a modular recorder for active or unattended bat monitoring.

Combines a Raspberry Pi-class computer, an ultrasonic microphone, local storage, and a web interface into a modular recorder for active or unattended bat monitoring.

### Why it matters {#why-it-matters-mp-2026-09-16-021}

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

### Limits and context {#limitations-mp-2026-09-16-021}

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

### Claims and sources {#claims-mp-2026-09-16-021}

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

## 24. OpenCTD {#mp-2026-09-16-022}

- Story ID: `mp-2026-09-16-022`
- Type: `invention_desk`
- Classification: `editorial`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-022/openctd

**Dek:** Packages conductivity, temperature, and depth sensors with an Arduino-compatible controller, battery, and SD storage inside a user-built housing for nearshore research and education.

Packages conductivity, temperature, and depth sensors with an Arduino-compatible controller, battery, and SD storage inside a user-built housing for nearshore research and education.

### Why it matters {#why-it-matters-mp-2026-09-16-022}

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

### Limits and context {#limitations-mp-2026-09-16-022}

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

### Claims and sources {#claims-mp-2026-09-16-022}

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

## 25. Microban {#mp-2026-09-16-023}

- Story ID: `mp-2026-09-16-023`
- Type: `invention_desk`
- Classification: `editorial`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-023/microban

**Dek:** Turns printable body parts, nineteen servomotors, a Raspberry Pi Zero 2 W, batteries, and shared control software into a compact humanoid platform for hands-on robotics work.

Turns printable body parts, nineteen servomotors, a Raspberry Pi Zero 2 W, batteries, and shared control software into a compact humanoid platform for hands-on robotics work.

### Why it matters {#why-it-matters-mp-2026-09-16-023}

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

### Limits and context {#limitations-mp-2026-09-16-023}

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

### Claims and sources {#claims-mp-2026-09-16-023}

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

## 26. The First Paid Slot {#mp-2026-09-16-024}

- Story ID: `mp-2026-09-16-024`
- Type: `invention_desk`
- Classification: `house_example`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-024/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-2026-09-16-024}

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

### Limits and context {#limitations-mp-2026-09-16-024}

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

### Claims and sources {#claims-mp-2026-09-16-024}

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

## 27. Put Your Project on the Desk {#mp-2026-09-16-025}

- Story ID: `mp-2026-09-16-025`
- Type: `invention_desk`
- Classification: `house_example`
- Content status: `carried_over`
- Permanent URL: https://themachinepress.com/story/mp-2026-09-16-025/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-2026-09-16-025}

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

### Limits and context {#limitations-mp-2026-09-16-025}

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

### Claims and sources {#claims-mp-2026-09-16-025}

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

## Normalized sources

- **source-2026-09-16-001:** [arXiv preprint 2609.16251](https://arxiv.org/abs/2609.16251) — arXiv; primary_research
- **source-2026-09-16-002:** [arXiv preprint 2609.16108](https://arxiv.org/abs/2609.16108) — arXiv; primary_research
- **source-2026-09-16-003:** [arXiv preprint 2609.16129](https://arxiv.org/abs/2609.16129) — arXiv; primary_research
- **source-2026-09-16-004:** [arXiv preprint 2609.16145](https://arxiv.org/abs/2609.16145) — arXiv; primary_research
- **source-2026-09-16-005:** [arXiv preprint 2609.16206](https://arxiv.org/abs/2609.16206) — arXiv; primary_research
- **source-2026-09-16-006:** [arXiv preprint 2609.16215](https://arxiv.org/abs/2609.16215) — arXiv; primary_research
- **source-2026-09-16-007:** [arXiv preprint 2609.16245](https://arxiv.org/abs/2609.16245) — arXiv; primary_research
- **source-2026-09-16-008:** [arXiv preprint 2609.16338](https://arxiv.org/abs/2609.16338) — arXiv; primary_research
- **source-2026-09-16-009:** [arXiv preprint 2609.16487](https://arxiv.org/abs/2609.16487) — arXiv; primary_research
- **source-2026-09-16-010:** [arXiv preprint 2609.16186](https://arxiv.org/abs/2609.16186) — arXiv; primary_research
- **source-2026-09-16-011:** [arXiv preprint 2609.16040](https://arxiv.org/abs/2609.16040) — arXiv; primary_research
- **source-2026-09-16-012:** [arXiv preprint 2609.16041](https://arxiv.org/abs/2609.16041) — arXiv; primary_research
- **source-2026-09-16-013:** [arXiv preprint 2609.16075](https://arxiv.org/abs/2609.16075) — arXiv; primary_research
- **source-2026-09-16-014:** [arXiv preprint 2609.16331](https://arxiv.org/abs/2609.16331) — arXiv; primary_research
- **source-2026-09-16-015:** [arXiv preprint 2609.16368](https://arxiv.org/abs/2609.16368) — arXiv; primary_research
- **source-2026-09-16-016:** [arXiv preprint 2609.16115](https://arxiv.org/abs/2609.16115) — arXiv; primary_research
- **source-2026-09-16-017:** [arXiv preprint 2609.16105](https://arxiv.org/abs/2609.16105) — arXiv; primary_research
- **source-2026-09-16-018:** [arXiv preprint 2609.16106](https://arxiv.org/abs/2609.16106) — arXiv; primary_research
- **source-2026-09-16-019:** [arXiv preprint 2609.16125](https://arxiv.org/abs/2609.16125) — arXiv; primary_research
- **source-2026-09-16-020:** [arXiv preprint 2609.16134](https://arxiv.org/abs/2609.16134) — arXiv; primary_research
- **source-2026-09-16-021:** [arXiv preprint 2609.16128](https://arxiv.org/abs/2609.16128) — arXiv; primary_research