Public wikis for LLMs.
Private spaces for your work.

What you can do with it

A law firm records the arguments that held and the ones that lost. A hospital department records its own case series, with who attests it and over how many cases. A robotics team records the conditions under which the sensor drifts, measured and dated. Knowledge that is worth something is usually not public: that is why you can register it by declaring who attests it and by what method — not only with a link.

Eight wikis we built with Sinapsi

They are the proof that it works, not the product. Yours is private unless you decide otherwise.

• The Agent Economy — What Agents Actually Pay Foragent-economyagent-paymentsai agentsap2interchangemcppaymentsstablecoinsverified-numbersx40251 pages on how AI agents pay and get paid, in two halves. The SETTLEMENT rail: the x402 protocol over HTTP and MCP, what regulators in eight jurisdictions have written, and a census of the paid-agent market enumerated three times by us — 14,766 listings on 3 August 2026 falling to a median of one call per listing by 24 August, while total calls dropped 19% in nineteen days. The AUTHORIZATION layer: Google AP2, Mastercard AP4M, Visa Intelligent Commerce, Stripe/OpenAI ACP and Coinbase agentic wallets, each graded on one question — has anyone outside the issuing company shipped a working implementation? Only one of the five can show named merchants. 1,552 claims checked against primary sources; 123 were removed for not surviving. Card interchange figures come from Visa's own 28-page schedule and 12 CFR 235, not from consultancy blogs. The enumeration scripts are included so anyone can contradict us.
• Coding — verified notes on debugging agent & web systemscodingllm agentsdebuggingreliabilitywebField-tested lessons — the observable symptom, the real cause, the fix, and what didn't work — for the ways LLM-agent and web stacks fail, often silently. Verified, with sources.
• CTF Cryptography — Recognize the Weakness, Reach the Attackcryptoctfsecurityrsa112 techniques for CTF crypto: given a signal, reach the right attack — with field-tested operational detail and a clear map of what does NOT work. Distilled from real CTF campaign notes, every claim cross-checked against the sources (2,442 checked, 50 refuted and corrected).
• Google Ads — Operate Without Burning Budgetgoogle-adsmarketingppcai agents31 pages for marketers and their AI agents: auction mechanics, campaign setup, policy traps that get accounts suspended, and what does NOT work. 1,110 claims checked against Google's official documentation — 99 refuted and corrected along the way.
• Model Context Protocol (MCP)mcpai agentsdeveloperssecurityConnect any client and build a secure remote MCP server: transports, auth (Bearer/OAuth), the five primitives, and the failure modes that actually bite. 21 pages, every claim checked against the sources — 72 refuted claims removed along the way.
• Problemi aperti e metodi verificatimatematicaquantumproblemi apertimetodo di ricerca275 pagine di matematica della ricerca: i problemi di Erdős, il problema PPT², i metodi che hanno funzionato e le rotte già esplorate e chiuse. Ogni pagina ha provenance e verifica; i pesi d'uso vengono dalla campagna reale.
• Previsione geopolitica — analisi generata automaticamentegeopoliticaprevisioneanalisimetodomachine-generatedAnalisi geopolitica generata da una macchina con metodo dichiarato: fatti datati e verificati sulle fonti, posizioni degli analisti classificate e segnate contro gli esiti, tesi con la loro frontiera di esclusioni, ancorate a base rate e teoria. Confidence e fonti indicate. Non è asserzione editoriale.
• Trading Edges — Documented, Costed, Refutedtradingfactor investingmarket anomaliestransaction costsreplication crisisquantitative financeMarket anomalies judged by evidence rather than by story. Each one carries its effect size, its proof regime (peer-reviewed, practitioner-attested, or measured by us), the transaction-cost threshold at which it stops paying, and what happened to it after publication. 947 claims were checked and 92 refuted and removed during verification. The pages saying a signal does NOT work are as load-bearing as the ones saying it does.
• US Politics — verified base rates and forecasting referenceus politicselectionsforecastingbase ratesmidterms 2026congresspollingReference-class base rates, polling methodology, institutional mechanics, and the 2026 races — the durable, source-verified knowledge for forecasting US elections and policy. Machine-generated, with primary sources and honest uncertainty.