DISCOVERY / THE OBSERVATION LAB

Where should we
look next?

Start with the angles you have. Explore another viewpoint. Find out which observation the model expects to help.

Synthetic Paris-area case

2 supplied tracks 22 angular samples

No target truth is sent to the planner.
OBSERVATION LAB / SUPPLIED ANGLESIllustrative district
Station 1Station 2LOCAL TOP VIEW · supplied station positions

Preparing the observation scene…

Supplied sight lines
Scenery, measurements and scale

The miniature district is an illustrative setting, not a map of Paris or a terrain model. Hide it to inspect the geometry alone. Data coordinates use equal kilometre scales in a local Earth-fixed frame. Optical station models are enlarged; their pins mark the supplied locations. Drawn ray lengths do not measure range. Replay advances supplied sample indices without generating new measurements.

01 / WHERE SHOULD WE OBSERVE NEXT?

Same horizon.
Different viewpoints.

Explicit model: independent new angular errors, exact stations and clocks, constant velocity in spherical ECEF. Visibility is geometric only.

Original angles and complete request

These synthetic observations are supplied explicitly. Candidate series contain times and expected precision, never invented measured angles.

{
  "observations": {
    "observers": [
      {
        "lat": 48.85,
        "lon": 2.35,
        "elevation_m": 35,
        "clock_offset_s": 0,
        "sigma_az_deg": 0.2,
        "sigma_el_deg": 0.2,
        "samples": [
          {
            "t_s": 0,
            "azimuth_deg": 101.991260935,
            "elevation_deg": 14.447682585
          },
          {
            "t_s": 10,
            "azimuth_deg": 93.467590277,
            "elevation_deg": 14.732612639
          },
          {
            "t_s": 20,
            "azimuth_deg": 84.786897815,
            "elevation_deg": 14.70245413
          },
          {
            "t_s": 30,
            "azimuth_deg": 76.336410848,
            "elevation_deg": 14.362974947
          },
          {
            "t_s": 40,
            "azimuth_deg": 68.446289854,
            "elevation_deg": 13.773863939
          },
          {
            "t_s": 50,
            "azimuth_deg": 61.324990456,
            "elevation_deg": 13.022637731
          },
          {
            "t_s": 60,
            "azimuth_deg": 55.051414611,
            "elevation_deg": 12.194843666
          },
          {
            "t_s": 70,
            "azimuth_deg": 49.607028444,
            "elevation_deg": 11.356342386
          },
          {
            "t_s": 80,
            "azimuth_deg": 44.917906873,
            "elevation_deg": 10.549458666
          },
          {
            "t_s": 90,
            "azimuth_deg": 40.888140057,
            "elevation_deg": 9.79700007
          },
          {
            "t_s": 100,
            "azimuth_deg": 37.420214884,
            "elevation_deg": 9.108372913
          }
        ]
      },
      {
        "lat": 48.85,
        "lon": 2.45,
        "elevation_m": 35,
        "clock_offset_s": 0,
        "sigma_az_deg": 0.2,
        "sigma_el_deg": 0.2,
        "samples": [
          {
            "t_s": 0,
            "azimuth_deg": 258.008739065,
            "elevation_deg": 14.447682585
          },
          {
            "t_s": 10,
            "azimuth_deg": 266.532409723,
            "elevation_deg": 14.732612639
          },
          {
            "t_s": 20,
            "azimuth_deg": 275.213102185,
            "elevation_deg": 14.70245413
          },
          {
            "t_s": 30,
            "azimuth_deg": 283.663589152,
            "elevation_deg": 14.362974947
          },
          {
            "t_s": 40,
            "azimuth_deg": 291.553710146,
            "elevation_deg": 13.773863939
          },
          {
            "t_s": 50,
            "azimuth_deg": 298.675009544,
            "elevation_deg": 13.022637731
          },
          {
            "t_s": 60,
            "azimuth_deg": 304.948585389,
            "elevation_deg": 12.194843666
          },
          {
            "t_s": 70,
            "azimuth_deg": 310.392971556,
            "elevation_deg": 11.356342386
          },
          {
            "t_s": 80,
            "azimuth_deg": 315.082093127,
            "elevation_deg": 10.549458666
          },
          {
            "t_s": 90,
            "azimuth_deg": 319.111859943,
            "elevation_deg": 9.79700007
          },
          {
            "t_s": 100,
            "azimuth_deg": 322.579785116,
            "elevation_deg": 9.108372913
          }
        ]
      }
    ]
  },
  "candidates": [
    {
      "id": "extend-west",
      "station": {
        "id": "observer_1",
        "lat": 48.85,
        "lon": 2.35,
        "elevation_m": 35
      },
      "sample_times_s": [
        110,
        120,
        130
      ],
      "sigma_az_deg": 0.2,
      "sigma_el_deg": 0.2,
      "cost_units": 1
    },
    {
      "id": "new-north",
      "station": {
        "id": "north",
        "lat": 48.92,
        "lon": 2.4,
        "elevation_m": 35
      },
      "sample_times_s": [
        110,
        120,
        130
      ],
      "sigma_az_deg": 0.2,
      "sigma_el_deg": 0.2,
      "cost_units": 2
    },
    {
      "id": "over-budget",
      "station": {
        "id": "east",
        "lat": 48.85,
        "lon": 2.5,
        "elevation_m": 35
      },
      "sample_times_s": [
        110,
        120,
        130
      ],
      "sigma_az_deg": 0.2,
      "sigma_el_deg": 0.2,
      "cost_units": 6
    }
  ],
  "objective": {
    "horizon_s": 140
  },
  "constraints": {
    "max_cost_units": 3,
    "min_elevation_deg": -2
  },
  "assumptions": {
    "independent_new_errors": true,
    "exact_stations_and_clocks": true,
    "constant_velocity": true
  }
}
03 / WHAT CHANGED?

Now add observations.
Does the forecast hold up?

The prediction did not contain new angles. This next call adds three fixed synthetic readings and fits all supplied observations again. They were generated independently, before any candidate was evaluated.

First compare the proposed viewpoints.The selected admissible series becomes available here.

This demonstration does not activate a real sensor. The new readings are explicitly marked synthetic; the server does not authenticate their origin.

Compare the prediction with new readings.

After choosing an admissible viewpoint, add its synthetic observations. Compare the original forecast with the refitted estimate: new readings can help, disagree, or make it less certain.

04 / TRY TO BREAK A DECLARED CLAIM

What if one track
is slightly wrong?

Apply a fixed set of hypothetical angle and clock changes to the east observer, then fit each case again. A counterexample challenges your stated stability limit; it does not explain the cause of a real observation.

Six explicit cases: unchanged control, ±azimuth bias, +elevation bias and ±clock correction. Same original tracks, same motion model, evaluation at 140 s. No new evidence is acquired.

A claim needs a boundary.

Declare a position-shift limit, run the six cases, and inspect the returned counterexamples.

THE SAME QUESTION, A DIFFERENT DOMAIN

What would make
this plan work?

Keep the essential work. Explore an extra few minutes or one optional task. Check the smallest supported change against a fresh simulation.

120 minutesThree synthetic tasks · no calendar data
Set your requirements

One extra minute costs 1 unit. Removing the preview costs 15 units of lost value. Building and checking are always required.

THE NOMINAL WORKLOAD120 min deadline
deadline
Build the feature60 min
Run the checks40 min
Polish the preview20 min
04080120160 min

Bar lengths show nominal minutes. The experiment applies the declared variation independently to each retained task.

Search, then check again.

All permitted changes are declared before sampling. One finalist receives a separately seeded confirmation.

DISCOVERY / METHOD LAB

Explore new methods.
See how they perform.

Rust generates and tests scoring rules within a fixed set of allowed operations. Inspect the formula, compare routes and check the independent validation.

METHOD RESEARCH / CLOSED EUCLIDEAN TOURSInside the experiment See gains, losses and recorded routes
  1. 01
    GenerateCompose bounded rules
  2. 02
    SimplifyKeep the same semantics
  3. 03
    TestCompare quality and cost
  4. 04
    ValidateChallenge a frozen rule
A SMALL LANGUAGE. AN OPEN SEARCH. FIXED RULE LANGUAGE
THE GEOMETRY

One rule. Every next step.

Preparing the interactive geometry…

Drag to orbit · scroll to zoom
Candidate Nearest neighbor + 2-opt
Illustrative geometry / no search result
YOUR EXPERIMENTGenerate a candidate.

Loading the declared training corpus…

Real computation. Reproducible seed.
The search returns a candidate and its full history. It does not establish an independent benefit.
THE SCIENTIFIC BOUNDARY

Invention is the beginning.
Evidence is another job.

01

Search on train.

Generate, mutate, and cross bounded expressions. Evaluate every candidate on the declared training cases. Preserve the attempts.

02

Freeze the method.

Lock the selected expression and its hash. Keep the evaluation procedure, compute budget, and baselines explicit.

03

Challenge it independently.

Evaluate a frozen candidate on held-out cases with a separate verifier. A failed test stays a failed test.

PUBLISHED INDEPENDENT CHECKS
STUDY 2 / COST-AWARE SEARCH

Loading the independent record.

Retrieving the published outcomes and acceptance checks. This record is separate from your live search.

No promotion claimedPredeclared empirical gates · no universal guarantee

Research scope: a scoring expression for closed, planar Euclidean tours. The language is bounded; it does not execute arbitrary code. Search does not establish novelty, optimality, or a benefit beyond the declared tests.

THE CAPABILITY LAB

A faster method has
to earn its place.

Compare it with the reference. Keep every test case. Review the complete outcome before making it available to an assistant.

01Candidate

An explicit implementation

02Study

Waiting for a complete result

03Review

Required before publication

04Available

Not active

Loading the published study record…

The model stays visible.

  • Conditional local linear forecast under a spherical constant-velocity model.
  • The original observations are fitted once; candidates add predicted information without invented angles.
  • Independent errors, exact clocks and station positions are explicit assumptions.
  • Terrain, weather, access and actual sensor availability are not established.
Call these tools independently through MCP