Perslis
Site map
Every page on perslis.com, nested by section. 156 pages
Shells · our first model and our shells 5 pages
- Perslis shells — and Peel, our first model
- Kist — the loop shell at the core of Perslis
- Perslis Symbols — ask a model once, keep a symbol that answers forever
- Perslis Symbols beta — send an export and your questions, get back a kit
- How Perslis Symbols works — the paper: spec not code, an admission gate, a person, a signed runtime
The Fail-First Model 4 pages
Symbolic AI 6 pages
- What Is Symbolic AI? Definition, How It Works, Examples
- The History of Symbolic AI: From Dartmouth to Neuro-Symbolic
- What Is Neuro-Symbolic AI? Hybrid AI Explained
- What Is a Symbolic System? Symbols, Rules and Reasoning
- What Are Symbolic Flows? Deterministic Pipelines for AI
- Symbolic AI — how Perslis verifies instead of trusting
Symbolic AI techniques 11 pages
- Symbolic AI Techniques: Every Major Method, 1956–Today
- Logic Programming, Prolog and Theorem Proving Explained
- Formal Verification and Program Synthesis Explained
- Search Algorithms in AI: A*, Minimax, Alpha–Beta, MCTS
- AI Planning: STRIPS, HTN, PDDL and Graphplan Explained
- Knowledge Representation in AI: Frames, Ontologies, OWL
- Nonmonotonic Reasoning and Commonsense Reasoning in AI
- Expert Systems: Production Rules, Rete, MYCIN and XCON
- Cognitive Architectures: Soar, ACT-R, Blackboards
- Constraint Satisfaction, SAT Solvers and SMT Solvers
- Symbolic Machine Learning: Decision Trees, ILP and More
Science · Peel 23 pages
- Peel — The Fail-First Model for science
- Airtight — your research stays on your machine
- Inference Placement — where ML earns a seat in Peel
- Peel Pilot Program — for scientists & labs
- Peel for bioscience — A foundation for research
- go solve lymphoma
- Lois research floor — live demo, nothing saved
- API
- Peel evidence — Public artifacts and verification
- How It Works
- Instruments
- Lois — your AI lab partner
- Meet the Model
- trial success predictor
- Peel accounts — privacy notice
- Product
- Reasoning — the symbolic core inside Lois
- The science runtime — Source-pinned bioinformatics tools with hands
- The Peel system — Knowledge, reasoning, and runtime
- Trust
- Product
- WetHands: bounded actuation and evidence-gated re-entry for the lab
- WetHands — one runtime for every tool in your lab
Science · the instrument wall 20 pages
- Tools — 44 bioinformatics research instruments on one console
- AlphaFold DB integration — predicted structures with pLDDT confidence
- arXiv search integration — quantitative-biology preprints for AI research
- bioRxiv & medRxiv integration — preprint feeds pinned to DOI
- ChEMBL integration — measured bioactivities against targets, pinned to source
- ClinVar integration — clinical significance reported verbatim, never re-graded
- dbSNP integration — rsID records pinned to location and alleles
- Ensembl integration — genes, coordinates and sequence pinned to source
- InterPro integration — protein families and domains, honest zeros
- KEGG integration — pathways, compounds, enzymes and drugs pinned to ID
- NCBI sequence integration — nucleotide and protein records pinned to accession
- OpenAlex integration — open scholarly metadata for AI research
- RCSB PDB integration — experimental protein structures pinned to source
- PubMed Central full-text integration — open-access papers by PMCID
- PubChem integration — compound identity pinned to CID and InChIKey
- PubMed integration for AI agents — source-pinned literature search
- Reactome integration — curated pathways and reactions with stable IDs
- Semantic Scholar integration — the forward citation graph for AI research
- UniProt integration for AI agents — source-pinned protein data
- Ensembl VEP integration — predicted variant consequences for research
Motion 11 pages
- Perslis Motion — intelligence proposes motion, the floor decides what can move
- Drive Floor — a runtime safety shield for autonomous vehicles
- The Perslis Motion engine — reachability, invariants, admission
- Flight Demo: the Floor at the Controls of a Drone
- Flight Floor — a runtime safety shield for drones and UAVs
- How the harness works — every step from proposal to actuation
- Perslis Motion for investors — the safety runtime under every autonomous system
- drive-floor · runtime safety shield
- Where the floor's rules come from — earned from the public failure record
- Receipts — auditable, certifiable autonomy
- Robot Floor — a runtime safety shield for robots and cobots
Defense 28 pages
- Adaptive autonomy, fixed authority
- Alignment: aligned to you, not to a model
- Any model on top — it only gets the allowed moves
- The legacy bridge — plugs into the gear you have
- Connection: every system you run, under one brain
- Pitch deck: adaptive autonomy, fixed authority
- Defense: stay alive, stay inside the lines
- VDSG demo: give it orders, watch it play by them
- Evidence — the wins and the losses
- Evolution: it writes its own rules from its failures
- Frozen weights or weightless learning: when to use which
- The brain factory — how the brain was built
- The embodied brain — how the brain was built
- Fail-First — how the brain was built
- Peel — a self-evolving symbolic brain — how the brain was built
- SML — the smallest language model — how the brain was built
- Structured knowledge — how the brain was built
- TinkyBrains — how the brain was built
- Trace — Peel — how the brain was built
- One brain, many worlds — how the brain was built
- History: how the brain was built, step by step
- The ladder — harder every rung, losses included
- Offense: it learns to win the fight
- Receipts: a written reason for every decision
- Robotics: from your systems to combat-capable humanoids
- The runtime — it keeps learning, never past your rules
- Squads — stack enough rules and it freezes
- Timeline: from a toy's brain to a combat humanoid under command
Research library 8 pages
- Research and white papers
- Perslis Arcade — Fail-First Models: failure becomes structure
- Fallout field notes — a fail-first pilot with a guidebook
- Watch a floor refuse — Space Invaders, no weights
- The floor vs the classifier — why a confidence score is not a check
- Solve a cube — the symbolic floor, stage by stage
- Findings ledger — every result and where it is written up
- Ask once. Own the answer. — Perslis Symbolic Floor
Perslis · the runtime and products 33 pages
- Perslis Machine Minds — Guessing isn’t
- Perslis Accessibility — AAC infrastructure for everyday AI
- The TinkySpeak API — embed it, or call it over HTTP
- Every TinkySpeak command — the console is the API with a keyboard
- Try TinkySpeak — watch a bilateral conversation build
- Get the TinkySpeak runtime
- Offline and privacy — a conversation that needs no connection
- Why we build it — bridging disability and everyday AI
- The TinkySpeak runtime — one conversation engine, any app on top
- See to Speak — a photograph becomes communication
- AirTight — the offline loop where nothing leaves your machine
- SYMBOLIC FLOOR — FIELD MANUAL (Rev. 2026.09)
- Benchmarks — how Perslis measures a body, not a head
- The Brain Factory
- Verified COBOL Synthesis — a no-bake, self-extending system (preprint)
- Perslis Pitch Deck
- Evolution — how no two Perslises end up alike
- Jira Monkey — a ticket, a model, a second look
- Legacy — reconstructing dead software and running it natively
- Liberate — legacy data recovery: verified bundles from the files nobody can open
- The Loop — how Perslis researches, decides, and builds
- Models — every one, on purpose: the curated roster Perslis drives
- Perslis Floor — exact answers from your own data, no model in the loop
- Perslis Software Preservation — run and preserve legacy Windows/DOS software
- Perslis Security privacy
- Perslis Security support
- Perslis Security terms and plans
- Perslis Systems — the AI operational runtime for your existing systems
- Tools — the 70+ Perslis ships with, and the live dashboard
- White paper — Perslis: the body that makes any model useful, proven from the binary
- Win95 USB Lab — why we built it, what we learned, and the full white paper
- Windows and USB
- YOLO mode — full autonomy, fail-closed