Every figure below is measured from a calibration run in
progress and will change as thresholds are revised. This is an instrumentation
dashboard, not the synt.fun product — the public interface is a separate build
not represented here.
Throughput
processed
—
tokens / hour
AI-scored
—
tokens / hour
selectivity
—
reach scoring
Cumulative totals since the engine last started.
99.1% of load eliminated before any paid call. Screening completes in the first 60 seconds using zero RPC calls.
Cost efficiency
RPC quota
—
—
AI spend
—
calls / hour · cap 60
decision latency
—
median, from token launch
headroom
—
capacity remaining
Model performance
—
Safety score calibration
score 0–14
2 tokens
15–27
3 tokens
28–44
7 tokens
45–64
4 tokens
65–100
0 tokens
The engine scores how safe a token is for a trader, not how far it may rise. Each cell shows the share of tokens in that band that did not collapse within 30 minutes, verified against realised price. A curve rising from left to right indicates the score is calibrated.
Awaiting verification. Outcome confirmation requires 30 to 60 minutes per token.
Historical accuracy
base rate
—
tokens that rise ≥3× unaided
screening lift
—
capture rate ÷ base rate
—
Proprietary dataset
tokens tracked
—
verified outcomes
creators profiled
—
cross-token history
wallets labelled
—
tracked occurrences
acquisition
—
tokens / day
now tracking
—
awaiting outcome
labelled / hour
—
verified vs price
verified / day
—
outcome confirmed
Operates independently of the screening engine. Every token is tracked until its outcome is verified against realised price. This dataset cannot be purchased or scraped.
Compounding asset. Cross-token creator history cannot be reconstructed from public data — it requires continuous observation from before each token exists. Coverage therefore compounds for each day the system remains in operation.
Reliability
component
status
note
engine
active
auto-restart
data feed
active
websocket
enrichment
active
0.5% error
scoring
active
0 timeouts
RPC error rate
—
—
metadata
—
of scored tokens
test suite
153
passing
labelled
—
this session
Live AI scoring feed
Scores are recorded before the outcome window opens and are never adjusted retroactively. Each entry becomes verified evidence after 30 minutes.
Engineering log
Each change that altered a reported figure, with its measured effect. Entries documenting our own errors are included by design: a system that reports no failures is not being measured.
Refreshed automatically every 15 seconds. All figures are read directly from the production database.
synt · autonomous screening engine for micro-cap markets. It processes every token launch in real time and produces calibrated safety scores, each verified against realised price.