Five labs work on one record of the Solana market.

Rating log16 Nov 2026
On a full day, 13% to 27% of ratings were High or Very high.
Every rating on the log from 2 Nov to 15 Nov, by class and day.
In 20 full days, the recording missed 3.7 minutes.
Market recording16 Nov 2026
Full report: method, limits, data
52% of all launches come from 1% of the devs.
Memory15 Nov 2026
Full report: method, limits, data
At 19:00 UTC, 1.9 times as many coins launch as at 07:00.
Market recording15 Nov 2026
Full report: method, limits, data
The units.
- 01
Market recording
Every launch, trade and pool change the hub reads, written down as it arrives.
11.6 GB of launch market dataRunning since 25 Oct 2026
2 reports
- 02
Memory
Each coin with its dev, its early buyers and what it did afterwards.
926,922 coins, 208,811 devsRunning since 25 Oct 2026
3 reports
- 03
Wallet index
Which wallets keep arriving early together. The ring index is rebuilt every 30 minutes.
2,240,554 wallets, 113,665,163 tradesRunning since 25 Oct 2026
None yet
- 04
Rating log
Every risk rating chained into a log the moment it is made, checked one hour later.
53,301 ratings to 15 Nov 2026Running since 2 Nov 2026
2 reports
- 05
Forecast register
A probability is shown only where a class beats the base rate on coins it never saw.
One class live, 4 in the gateRunning since 18 Nov 2026
None yet
Counted on the production hub on 15 Nov 2026. Units in build are not live yet and hold no counted data.
Four rules hold for every unit.
- Checked after one hour
- A rating counts only once the coin's market cap one hour later is on the log, whatever it shows.
- Holdout before probability
- A forecast class gets a number only after it beats the base rate on coins held out of its training.
- Own data only
- Third-party sources (pump.fun, RugCheck, DexScreener, RPC providers) are read and turned into CON's own fields. Their raw answers are never passed on.
- Every report names its limits
- Method and limits stand under every finding, so a reader can check what the number does not show.
Open questions.
What the units work on next. A question gets an answer only once it is counted, never before.
- MemoryHow do coins end since the pull detection was fixed?
- MemoryDo the coins of serial devs go less far than the coins of one-time devs?
- Rating logDoes the result of a rating depend on the market cap at the moment of the rating?
- MemoryIs the share of coins that never pass $5,000 the same from week to week?
- Forecast registerWhich risk class beats the base rate on coins held out of its training?
What the labs publish.
- ResearchReports with their figures, method and limits.
- Model cardsEvery forecast class with its holdout test, skill, base rate and gate state, and what did not work.
- Market dataThe launch market per day, with the gaps in the recording.
- MethodologyHow the market is recorded, the words every report uses and the known weaknesses.
- CorrectionsEvery change to data and reports, with its date.
- Rating logEvery rating with its outcome after one hour, hits and misses, with a daily closing hash.