Desks and funds
Your scanner half works and the last dev shop is gone. We build venue integrations, matchers, and risk rails you own outright, spec-first and test-gated.
What a build includesCONNECTING VENUE FEEDS
The arbitrage machine that shows its refusals.
Scanning Polymarket, Kalshi, and Hyperliquid around the clock. The math decides, and it trades only what it can prove.
24/7scan loop
3venues
869tests
0LLMs in the trade loop
0days in shadow
0opportunities priced
0decisions on record
0refusals written down
0order books archived
0funding snapshots
0cross-venue pairs under watch
0automated tests
It is deciding right now.
This is the scanner's actual output, exported from the decision ledger every three minutes: every opportunity priced, the edge the arithmetic computed, and the verdict. Detections and verdicts only, never positions. It has run every hour of every day since August 12.
…opportunities priced, 24h
…refusals journaled, 24h
…cross-venue pairs watched
…since the last decision
DECISION STREAMshadow mode · auto-refreshes
- …connecting to the decision ledger
Built for people losing money to bad edges.
Basis is not a fund and sells no signals. It is a working system, and the quantitative process that built it is for hire: priced fee curves, strict matching logic, sizing against observed depth. No narrative, no dream.
Founders of market products
Venue connectors and risk tooling shipped against a written spec in weeks, with a test suite and runbook your engineers inherit cleanly.
Scope and timelinesOperators of leaking bots
Fills look fine, money still leaks. The audit runs your pairs and fills through the failure catalog and names every trap, with the cost of each.
What the audit findsSCAN
Every book, every venue, all day.
Polymarket slates, the full windowed Kalshi universe, and hourly funding on every Hyperliquid perp. Budgeted polling with failure accounting, so an empty lane can never be silent.
3venues
722,921order books archived
232perps on the funding field
MATCH
The same question is not the same market.
Each market's settlement rules are fingerprinted: sources, dates, operators, thresholds. A pair is tradeable only when every piece matches, and none has cleared that bar yet. The bar is the product.
308pairs held at likely
0rated exact
1 in 3published pairs differ
REFUSE
The default answer is no.
Implausible edges are treated as bait. Sub-rail edges are treated as noise. Non-exact pairs are treated as evidence. The biggest edge this machine has ever seen, it turned down.
435refusals journaled
10.95%biggest edge refused
$0.02fill rail per set
JOURNAL
Every decision, written down.
Fills, skips, venue closures, rule edits: one ledger holds all of it. Capital unlocks only when that ledger clears evidence gates, and it re-locks the moment live results drift from the record.
3,451decisions on record
3money gates
1ledger, no edits
It found a 10.95% edge. Then it said no.
Two venues, one bitcoin question, a set that costs well under a dollar and pays $1. On paper, double-digit locked-in return in four days. The machine journaled it, refused it capital, and moved on. The mismatch was the edge.
buy YESpolymarket:3763328 @ 0.0570
buy NOkalshi:KXBTCD-26AUG2817-T80499.99 @ 0.6000
paircost $0.9008, edge $0.0992, tier likely
verdictREFUSED · rules-differ, threshold-differs
Money unlocks in stages.
Basis runs a full simulated book against real prices, real fees, and real depth. It earns its way to live capital by proving itself on the record, not by a good week. These gates govern the operator's own research capital only: Basis is not a fund and takes no outside money.
$0
Shadow mode. Prove 1% per month net over 20 closed fills. Running now.
$100
First live capital. Plumbing built, tested, and deliberately disarmed.
$500+
Scales only if live results track shadow within 30%. Otherwise it stays small.
Work with the people who built it.
Trading infrastructure builds
Scanners, matchers, execution rails, and honest dashboards for your venue or strategy. Designed, tested, and shipped the way Basis was.
from $10,000
Request a build
Strategy risk audit
Send your bot's pairs and fills. We run them through the failure catalog and name every leak, with the estimated cost of each.
$2,500 flat
What the audit finds
Every engagement ships code you own outright, with the written spec, tests, and runbook to run it without us. Emails get a reply within one business day.