Perception
Computer vision
Perception that degrades gracefully — 3D reconstruction, detection and tracking tuned for the conditions a demo never shows you.
Computer vision · Language models · Robotics
Most AI demos are screenshots. Ours run in your browser.
Entroπa Labs is an independent AI practice. Ten working systems — radiance fields, model internals, a robot sandbox, a forecast that grades itself nightly, an AI detector and an image studio — each one open on this site, with the mechanism written down and the limitations left in.
What we do
Three fields, one discipline: make it hold outside the lab.
Benchmarks reward the average case. The world serves the tail — drift, noise, edge conditions. We work where models meet reality, and measure what survives contact.
Language
Language models
LLM systems you can inspect and trust — routing, evaluation, and the plumbing that turns a capable model into a dependable one.
Embodiment
Robotics
Control and planning that close the loop under uncertainty — policies that recover when the world refuses to cooperate.
Writing
One post per system, from the source.
Not announcements — explanations. Each one takes a single application apart: the mechanism underneath, the part that turned out to be harder than it looked, and the limitations left in rather than tidied away.
9 postsNewest first
every one open to inspection
15 Aug 2026Detecting generated media without a black boxSix stylometric tells, a forensic chain of custody for pixels, and a gauge that refuses to say 100% — how the AI detector works, and the paragraph on what it cannot prove.08 Aug 2026Every AI deadline, and the ones nobody publishesMerging two volunteer feeds, destroying every date string on sight, and the discipline of leaving a column blank when the honest answer is that nobody published it.26 Jul 2026The agent loop, and what it can't do in a browserA tool-use agent is about forty lines of control flow. Everything difficult about it lives in the tools you are willing to hand it.25 Jul 2026Reading a checkpoint in the browsersafetensors is a length, some JSON and a wall of bytes. A torch .pt is a ZIP with a pickle in it — which means writing a small pickle VM.25 Jul 2026Deciding which experts to dropMost of an MoE's experts are underused and several are near-duplicates. Turning that into a defensible pruning plan needs two numbers, not one.25 Jul 2026A 3D renderer in 120 lines of canvasNo WebGL, no matrices library, no dependencies: project, cull, shade, sort. The robot playground draws every frame the way graphics worked before GPUs.25 Jul 2026A forecast that grades itselfAnyone can publish probabilities. The interesting question is whether they mean anything — so this one scores itself every night and refits.25 Jul 2026Valuing a hockey player from a box scorePoints tell you who is good. Surplus value tells you who is *tradeable* — and the gap between those two is where a GM actually works.24 Jul 2026How the Gaussian splatting viewer worksA radiance field is millions of fuzzy ellipsoids. Getting them on screen at 60fps is mostly a sorting problem, and the sort is the interesting part.
All writing Our research
Ten working systems, each one open to inspection.
Our research ships as software you can run, not screenshots. Each entry below is a live application — the purpose it was built for, the mechanism underneath, and the door into it. Most run entirely in your browser; the two generative studios run on our own edge — either way, nothing you upload is stored on a server.
10 / 10 liveAcross three practices
generative · agentic · physical
Generative AI
The models themselves — opening a checkpoint, reading the weights, deciding which experts earn their parameters, driving one text-first, and telling their output apart from ours.
4 systemsAgentic AI
Systems that decide and act unattended — a tool-use loop, a forecast that grades itself nightly, and a decision engine over live data, and a deadline board that maintains itself.
4 systemsPhysical AI
Systems tied to the physical world — reconstructing a captured space, and driving a body through a simulated one.
2 systems
ParallaxPowered by Entroπa Labs
Physical AIPerception · 3D reconstruction
Live
Gaussian Splatting Viewer
PurposeJudge a radiance-field reconstruction by moving through it. A 3DGS scene looks flawless from the training views and falls apart two metres to the left; this puts free viewpoint control in front of the reconstruction so the failure shows itself.
Function
- Loads a .splat or an Inria .ply straight off disk
- Packs every gaussian's covariance into an integer data texture, renders with instanced billboards
- Depth-sorts every frame in a Web Worker so the main thread stays free
- Orbit, zoom, shift-pan, and a splat-size slider; a demo scene is embedded so it is never empty
StrataPowered by Entroπa Labs
Generative AILanguage models · Interpretability
Live
Weights Visualiser
PurposeMake a checkpoint legible. A model file is usually a black box you load and hope about. This one opens it in the browser and shows the architecture, the actual weight values, and where the tokens go.
Function
- Parses .safetensors in full; torch .pt/.pth best-effort via a ZIP reader and a minimal pickle VM
- Architecture chart of parameters per layer, log-scaled and typed by colour
- Weight matrix as a diverging heatmap, plus a value histogram with min/max/mean/std/%≈0
- Token routing — a simulated forward pass animating each token to its top-k experts with live per-expert load
QuorumPowered by Entroπa Labs
Generative AILanguage models · Efficiency
Live
Mixture-of-Experts Pruner
PurposeDecide which experts you can drop, and say what it costs. Most of an MoE's experts are not pulling their weight and several are near-duplicates of each other. This makes that argument with numbers instead of intuition.
Function
- Utilisation — share of routed tokens per expert, from a seeded router simulation
- Redundancy — cosine similarity between down-sampled per-expert weight signatures
- Keep-score ranking: normalised utilisation minus a redundancy penalty
- A live pruning plan — experts cut, parameters saved, and a clearly labelled heuristic retention estimate; downloadable as a standalone report
StridePowered by Entroπa Labs
Physical AIRobotics · Embodiment
Live
Robot Playground
PurposeBuild intuition for embodied control with nothing installed. Morphology and terrain decide as much as the controller does; here you can change both in a second and feel the difference on the keyboard.
Function
- Five bodies — humanoid, quadruped, rover, robot arm, drone — built from nested articulated frames
- Five worlds — studio, grid arena, obstacles, ramps, low-gravity moon
- Procedural gaits driven live: WASD to move, space to jump, R/F for altitude, per-joint keys for the arm
- A dependency-free software 3D renderer on a 2D canvas: project, backface-cull, flat-shade, painter's sort
HorizonPowered by Entroπa Labs
Agentic AIAggregation · Public record
Live · daily
AI conference deadlines
PurposePut every submission date a researcher needs in one place, and keep it honest. The deadline lists the field relies on are maintained by volunteers in several places at once; the interesting engineering is reconciling them, resolving AoE into a real instant, and saying so out loud when two sources disagree.
Function
- Abstract, paper, rebuttal, notification and camera-ready dates for the major ML, vision, language, robotics and data-mining venues
- Two maintained feeds merged and cross-checked — a paper deadline that differs by more than six hours is flagged on the row
- Workshops, accepted-paper counts, orals and spotlights read from OpenReview's public API and cached permanently, so the record deepens daily
- A subscribable .ics feed at a stable URL, plus registration fees read from each conference's own page with the date they were read
VerityPowered by Entroπa Labs
Generative AIForensics · Provenance
Live
AI Detector
PurposeTell how likely a piece of text, an image, or a video is machine-made — and show every reason. Commercial detectors return a bare percentage and keep your files; this one runs entirely in the browser and decomposes its verdict into named, arguable signals.
Function
- Text stylometry — sentence-rhythm burstiness, stock-LLM phrasing, phrase recycling, vocabulary spread, absence of human mess, essay-template structure — with the most AI-flavored passages highlighted
- Image forensics — Content Credentials / C2PA manifests, embedded generator metadata, camera EXIF provenance, diffusion-grid geometry, and block-wise sensor-noise statistics
- Video — container markers plus frame-sampled noise analysis
- The gauge saturates at “very likely”, never “certain” — a likelihood with reasons, not a verdict to punish someone with
GlyphPowered by Entroπa Labs
Generative AIImage generation · Typography
Live
AI Image Studio
PurposeMake image generation feel like a studio, not a slot machine. Legible text inside an image is a first-class control with its own field; styles, palettes and canvases are explicit dials; and every result lands on a persistent pinboard in your browser rather than vanishing into a chat scroll.
Function
- An exact-text field that steers the model to render your words legibly, spelling intact — typography as a control, not a lucky roll
- Ten style presets, seven palettes, three canvases, quality and batch dials — assembled into one engineered prompt
- ✨ Magic rewrite polishes a rough idea into a vivid prompt with a text model before the image model sees it
- A masonry pinboard with search, pinning, remix, download and lightbox — stored in IndexedDB, so the board is yours and stays on your machine
- Photo studio — upload a photo and restyle it into a pack of dating, professional or studio looks; keeps your pose and vibe (not an exact-face clone), no watermarks
- Free built-in engine — FLUX running on the site's own edge, a free Google sign-in and no API key, with a fair-use daily allowance; an optional bring-your-own-key engine unlocks bigger canvases
PriorPowered by Entroπa Labs
Agentic AIApplied forecasting · Calibration
Live · daily
Beat the Streak
PurposeRun a probability model in public and grade it every day. Calibration is the only claim that matters and the only one most models never publish. Baseball supplies a fresh, honest test slate every morning.
Function
- Ranks every likely starter by modelled P(at least one hit) for today's slate
- Features: matchup, contact and plate skills, recent form, platoon split, ballpark, batter-vs-pitcher history
- Rebuilt each morning by a stdlib-only Python pipeline on a scheduled workflow
- Yesterday's board is graded against results and kept as history, so the weights answer for themselves
LedgerPowered by Entroπa Labs
Agentic AIDecision support · Valuation
Live · daily
NHL GM Playbook
PurposeTurn public data into the decision a general manager actually faces. Not "who is good" but who is underpriced, who is movable, what a team is missing, and which side wins the trade.
Function
- Scout — breakout score, box-score WAR and $/WAR, and surplus trade value for every qualified skater
- Contract projections — next-deal AAV and term, live from CapWages with runtime anchor validation and a CSV fallback
- Team Lab, all 32 clubs — optimal lines and handedness-balanced pairs, starter, roster needs, movable players, farm pipeline, fair trade ideas
- Trade analyzer — build both sides and get a value verdict on who wins
ParleyPowered by Entroπa Labs
Agentic AILanguage models · Agents
Live
Voice Agent
PurposePut the agent loop somewhere you can watch it work. Tool use is usually described in a diagram; here you ask a question out loud and see the model choose a tool, the tool run against real data, and the answer come back — the whole cycle, in the open.
Function
- Speech in, speech out — Web Speech recognition with a live level meter, answers spoken back; typing works everywhere
- Two provider adapters, the only two wire formats there are: Anthropic content blocks and the OpenAI-compatible tool-calls array
- Tools query this site's own live data — today's hit probabilities, skater valuations, a club's roster needs
- Your key, your browser: kept in
localStorageand sent only to the provider — there is no server here to send it to
Each system has a written companion on the blog — what it is for, the mechanism underneath, and the part that was harder than it looked. Papers land here as the work produces them; the shelf is deliberately empty rather than padded. The systems above are the current record: reproducible by construction, offline by default, and honest about which figures are measured and which are estimates.
Pricing
Free to start. Pay only for the heavy lifting.
Every app's core is free forever and runs in your browser. Subscriptions unlock the compute-heavy extras — more image generations, photo packs, bigger canvases, PDF & batch analysis, exports and priority.
Free
$0
Full AI detector, all demos, 5 studio images/day, local pinboard & workspaces.
Pro
$9/mo
100 images/day, photo packs, upscale & background removal, batch ZIP export, PDF detection, downloadable reports & scan history.
Studio
$19/mo
500 images/day, highest priority, largest batches, early access to new modules.
Contact
Have a system that has to hold in the field?
hello@General enquiries, and anything that does not fit below
collaborate@Engagements, partnerships and joint research
recruiting@Roles, introductions and speculative applications
support@Something on this site is broken or wrong
Independent AI practice · computer vision, language models and robotics · assessments and hands-on engagements.