Rebuild org card as pure markdown: banner, badges, honesty table, links as bullets

README.md

@@ -1,58 +1,80 @@
-## Dhi Labs
+![Dhi Labs: edge vision AI that refuses to guess](banner.svg)
 
-**Edge vision AI that refuses to guess: calibrated, auditable, honest by construction.**
+![products](https://img.shields.io/badge/products-6-334155)
+![demo spaces](https://img.shields.io/badge/demo%20spaces-6-0ea5e9)
+![benchmarks](https://img.shields.io/badge/benchmarks-gated-f59e0b)
+![code](https://img.shields.io/badge/code-closed%20source%2C%20open%20evidence-111827)
 
-Most edge-vision systems answer every query with false confidence: a bare count, a bare
-detection, a bare alert, with no sense of when the system does not actually know. Dhi Labs
-builds the opposite: components that quantify their own uncertainty, ship a falsification
-ledger instead of a marketing claim, and say "I don't know" instead of guessing when they are
-outside their calibrated support.
+## Who we are
 
-Every artifact on this org states plainly whether its numbers are synthetic or real-world, and
-we never claim state-of-the-art or foundational status, only what was actually measured,
-against which ground truth, and what has not been checked yet.
+Dhi Technologies builds video analytics software that runs directly on Jetson class edge
+hardware, close to the camera, rather than shipping frames to a cloud model. Most edge vision
+systems answer every query with false confidence: a bare count, a bare detection, a bare alert,
+with no sense of when the system does not actually know. Dhi Labs builds the opposite:
+components that quantify their own uncertainty, ship a falsification ledger instead of a
+marketing claim, and say "I don't know" instead of guessing when they are outside their
+calibrated support.
 
-### Products (each with a live static demo Space)
+## Honesty, as a feature
 
-| Product | What it does | Demo |
-|---|---|---|
-| **[Amodal Counting](https://huggingface.co/datasets/Dhi-Technologies/amodal-counting-benchmark)** (A4) | Visibility-corrected counting through crowds and occlusion: a calibrated interval instead of a bare point count. | [Space](https://huggingface.co/spaces/Dhi-Technologies/amodal-counting-demo) |
-| **[Multicam Reasoning Memory](https://huggingface.co/datasets/Dhi-Technologies/multicam-reasoning-memory-benchmark)** (E1) | Cross-camera identity linking with transit-time priors and weeks-scale bounded memory; every answer carries provenance, and it refuses to link when unsure. | [Space](https://huggingface.co/spaces/Dhi-Technologies/multicam-reasoning-memory-demo) |
-| **[Causal Predictive Alerting](https://huggingface.co/datasets/Dhi-Technologies/causal-predictive-alerting-benchmark)** (E4) | Predicts an incident seconds before it happens from kinematic trajectories, and proves *why* via counterfactual replay and a fulfilled/falsified ledger. | [Space](https://huggingface.co/spaces/Dhi-Technologies/causal-predictive-alerting-demo) |
-| **[Fixed-Camera 3D](https://huggingface.co/datasets/Dhi-Technologies/fixed-camera-3d-benchmark)** (A3) | Turns an ordinary fixed camera into a metric 3D sensor by self-calibrating from people walking through the scene: no GPU, no model weights. | [Space](https://huggingface.co/spaces/Dhi-Technologies/fixed-camera-3d-demo) |
-| **[Thermal Perception](https://huggingface.co/datasets/Dhi-Technologies/thermal-perception-benchmark)** (A5) | A radiometric data engine and self-supervised pretraining harness for thermal-native perception, with CPU-verifiable math ahead of any GPU pretraining run. | [Space](https://huggingface.co/spaces/Dhi-Technologies/thermal-perception-demo) |
-| **[Prompt2Model](https://huggingface.co/datasets/Dhi-Technologies/prompt2model-examples)** (B1) | A language-guided vision model factory: prompt to dataset to trained model to a calibrated, quantized, ONNX-exported artifact, with an accuracy-floor refusal gate. | [Space](https://huggingface.co/spaces/Dhi-Technologies/prompt2model-demo) |
+| Mechanism | What it means |
+|---|---|
+| **Refusal gates** | Components decline to answer when the input falls outside their calibrated support, instead of emitting a confident guess. |
+| **Calibrated confidence** | Where a product reports an interval or a confidence score, the coverage of that interval against ground truth is itself measured and reported, including when it is imperfect. |
+| **Falsification ledgers** | Predictive components log what they predicted, what actually happened, and whether the prediction was fulfilled or falsified, rather than only surfacing hits. |
+| **Stated limitations** | Every dataset and demo states plainly whether its numbers are synthetic or real world, and what has not been checked yet. No claims of state of the art or foundational status, only what was actually measured. |
 
-### Honesty, as a feature
+Two more things follow from the same policy: no SOTA claims, no customer names, and no
+deployment claims (this org is a public research and engineering surface, not a sales page), and
+every benchmark below is synthetic first and disclosed as such, so every claim can be checked
+exactly while real world validation is still in progress.
 
-- **Synthetic-first, disclosed as such.** Every dataset above is built from procedurally
-  generated ground truth so every claim can be checked exactly, and every card says so, plus
-  states plainly that **real-world validation is still in progress**.
-- **No SOTA claims, no customer names, no deployment claims.** This org is a public research and
-  engineering surface, not a sales page.
-- **Calibration is measured, not assumed.** Where a product reports an interval or a confidence,
-  the coverage of that interval against ground truth is itself a reported number, including when
-  it is imperfect.
+## Products and evidence
 
-### Research status
+Six products, each with a gated benchmark dataset built from procedurally generated ground truth
+and a live static demo Space you can exercise in the browser.
+
+| Product | What it does | Demo | Dataset |
+|---|---|---|---|
+| **Amodal Counting** (A4) | Visibility corrected counting through crowds and occlusion: a calibrated interval instead of a bare point count. | [Space](https://huggingface.co/spaces/Dhi-Technologies/amodal-counting-demo) | [Dataset](https://huggingface.co/datasets/Dhi-Technologies/amodal-counting-benchmark) |
+| **Multicam Reasoning Memory** (E1) | Cross camera identity linking with transit time priors and weeks scale bounded memory; every answer carries provenance, and it refuses to link when unsure. | [Space](https://huggingface.co/spaces/Dhi-Technologies/multicam-reasoning-memory-demo) | [Dataset](https://huggingface.co/datasets/Dhi-Technologies/multicam-reasoning-memory-benchmark) |
+| **Causal Predictive Alerting** (E4) | Predicts an incident seconds before it happens from kinematic trajectories, and proves why via counterfactual replay and a fulfilled or falsified ledger. | [Space](https://huggingface.co/spaces/Dhi-Technologies/causal-predictive-alerting-demo) | [Dataset](https://huggingface.co/datasets/Dhi-Technologies/causal-predictive-alerting-benchmark) |
+| **Fixed Camera 3D** (A3) | Turns an ordinary fixed camera into a metric 3D sensor by self calibrating from people walking through the scene: no GPU, no model weights. | [Space](https://huggingface.co/spaces/Dhi-Technologies/fixed-camera-3d-demo) | [Dataset](https://huggingface.co/datasets/Dhi-Technologies/fixed-camera-3d-benchmark) |
+| **Thermal Perception** (A5) | A radiometric data engine and self supervised pretraining harness for thermal native perception, with CPU verifiable math ahead of any GPU pretraining run. | [Space](https://huggingface.co/spaces/Dhi-Technologies/thermal-perception-demo) | [Dataset](https://huggingface.co/datasets/Dhi-Technologies/thermal-perception-benchmark) |
+| **Prompt2Model** (B1) | A language guided vision model factory: prompt to dataset to trained model to a calibrated, quantized, ONNX exported artifact, with an accuracy floor refusal gate. | [Space](https://huggingface.co/spaces/Dhi-Technologies/prompt2model-demo) | [Dataset](https://huggingface.co/datasets/Dhi-Technologies/prompt2model-examples) |
+
+Everything above, plus the blog dataset, is indexed in one place: the
+[Dhi Labs collection](https://huggingface.co/collections/Dhi-Technologies/dhi-labs-honest-edge-vision-ai-6a4eb297cbd60f5f673cc2d7).
+Browse [all datasets](https://huggingface.co/Dhi-Technologies?type=dataset) or
+[all Spaces](https://huggingface.co/Dhi-Technologies?type=space) directly.
+
+## Writing (the blog)
+
+Four technical posts, each grounded in the committed repo numbers. Hugging Face has no public
+Posts or Articles API, so these live as versioned markdown in the
+[blog dataset](https://huggingface.co/datasets/Dhi-Technologies/blog).
+
+- **[AI that refuses to guess](https://huggingface.co/datasets/Dhi-Technologies/blog/blob/main/01_ai_that_refuses_to_guess.md)**: the thesis, calibration, refusal gates, provenance, and falsification ledgers as product features.
+- **[Precision first cross camera linking (E1)](https://huggingface.co/datasets/Dhi-Technologies/blog/blob/main/02_e1_precision_first_linking.md)**: how a uniqueness guard lifted synthetic site precision from 0.918 to 1.0, and how weeks scale memory stays under 2 MB.
+- **[When the error bar is the product (A4)](https://huggingface.co/datasets/Dhi-Technologies/blog/blob/main/03_a4_calibration_over_accuracy.md)**: calibration honesty, the analytic interval undercovered (0.50 to 0.63 versus a claimed 0.90); conformal calibration widened it 6.7x to 14.5x.
+- **[Six products, one honesty thesis](https://huggingface.co/datasets/Dhi-Technologies/blog/blob/main/04_portfolio_overview.md)**: a portfolio overview tying all six products to the honest by construction thesis.
+
+## Research status
 
 Research papers describing these methods are in preparation and have not yet been published on
 arXiv or any other venue. The two whitepapers on the labs page are internal architecture write
 ups, not peer reviewed papers, and are labeled as such.
 
-### Writing (the blog)
-
-Four technical posts, each grounded in the committed repo numbers:
-
-1. **[AI that refuses to guess](https://huggingface.co/datasets/Dhi-Technologies/blog/blob/main/01_ai_that_refuses_to_guess.md)**: the thesis, calibration, refusal gates, provenance, and falsification ledgers as product features.
-2. **[Precision first cross camera linking (E1)](https://huggingface.co/datasets/Dhi-Technologies/blog/blob/main/02_e1_precision_first_linking.md)**: how a uniqueness guard lifted synthetic site precision from 0.918 to 1.0, and how weeks-scale memory stays under 2 MB.
-3. **[When the error bar is the product (A4)](https://huggingface.co/datasets/Dhi-Technologies/blog/blob/main/03_a4_calibration_over_accuracy.md)**: calibration honesty, the analytic interval undercovered (0.50 to 0.63 vs a claimed 0.90); conformal calibration widened it 6.7x to 14.5x.
-4. **[Six products, one honesty thesis](https://huggingface.co/datasets/Dhi-Technologies/blog/blob/main/04_portfolio_overview.md)**: a portfolio overview tying all six products to the honest-by-construction thesis.
+## Code and evidence
 
-(Hugging Face has no public Posts/Articles API, so these live as versioned markdown in the
-[blog dataset](https://huggingface.co/datasets/Dhi-Technologies/blog).)
+The code itself is proprietary and closed source permanently. These gated datasets, benchmarks,
+and demo Spaces are the published evidence: closed source code, open evidence. Prompt2Model is a
+deliberate exception with one public, MIT licensed release: [Prompt2Model, Language Guided Vision
+Model Factory](https://github.com/DHI-Technologies-Inc/Prompt2Model-Language-Guided-Vision-Model-Factory).
+The GitHub org linked below is an identity link, not a browsable destination for the other five
+products.
 
-### Links
+## Links
 
 - **Website:** [dhi-tech.com](https://dhi-tech.com)
 - **Labs (live page, two whitepapers):** [dhi-tech.com/labs](https://dhi-tech.com/labs)
@@ -62,7 +84,7 @@ Four technical posts, each grounded in the committed repo numbers:
 - **Blog:** [Dhi Labs technical posts](https://huggingface.co/datasets/Dhi-Technologies/blog)
 - **Code:** proprietary, closed source permanently. These datasets, benchmarks, and demo Spaces
   are the published evidence; the code itself is not released. Prompt2Model is a deliberate
-  exception with one public, MIT-licensed release: [Prompt2Model, Language Guided Vision Model
+  exception with one public, MIT licensed release: [Prompt2Model, Language Guided Vision Model
   Factory](https://github.com/DHI-Technologies-Inc/Prompt2Model-Language-Guided-Vision-Model-Factory).
   The GitHub org [github.com/DHI-Technologies-Inc](https://github.com/DHI-Technologies-Inc) is an
   identity link, not a browsable destination for the other five products.

style.css

@@ -1,83 +0,0 @@
-body {
-	padding: 2rem;
-	max-width: 900px;
-	margin: 0 auto;
-	font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", "Arial", sans-serif;
-	line-height: 1.55;
-}
-
-h1 {
-	font-size: 28px;
-	margin-top: 0;
-}
-
-h2 {
-	font-size: 19px;
-	margin-top: 2rem;
-}
-
-h3 {
-	font-size: 15px;
-	margin-bottom: 4px;
-}
-
-p {
-	color: rgb(70, 76, 88);
-	font-size: 15px;
-	margin-bottom: 10px;
-	margin-top: 5px;
-}
-
-hr {
-	border: none;
-	border-top: 1px solid rgb(230, 230, 230);
-	margin: 1.5rem 0;
-}
-
-table {
-	border-collapse: collapse;
-	width: 100%;
-	margin: 0.75rem 0 1rem;
-}
-
-th, td {
-	border: 1px solid rgb(225, 225, 225);
-	padding: 10px 12px;
-	vertical-align: top;
-	text-align: left;
-}
-
-th {
-	background: rgb(248, 248, 250);
-}
-
-a {
-	color: rgb(37, 99, 235);
-	text-decoration: none;
-}
-
-a:hover {
-	text-decoration: underline;
-}
-
-@media (prefers-color-scheme: dark) {
-	body {
-		background: rgb(14, 16, 20);
-		color: rgb(220, 222, 226);
-	}
-	p {
-		color: rgb(170, 175, 185);
-	}
-	hr {
-		border-top: 1px solid rgb(45, 48, 55);
-	}
-	th, td {
-		border: 1px solid rgb(50, 53, 60);
-	}
-	th {
-		background: rgb(24, 26, 32);
-	}
-	a {
-		color: rgb(96, 150, 250);
-	}
-}