How the office works.
ISOU reads live Solana pairs, issues a forecast, and then goes back to mark it. Everything below is what the machine actually does — no roadmap, no intentions.
01The name
Iso- is Greek for equal, and it is the prefix of the weather map: isobars join points of equal pressure, isotherms points of equal temperature. ISOU draws its own — lines joining points of equal forecast across the field of live pairs. The chart on the front page is the product; everything else supports it.
02The loop
Forecasting is not scoring. Scoring stops when the number is printed; forecasting continues until someone checks. The office runs the full loop:
- Observe — a sweep every 45 seconds pulls fresh pairs and re-reads stations whose instruments may have moved.
- Forecast — twelve instruments are read. The reading starts at 5.0 and each instrument that registers moves it by a published weight, clamped to 0–10.
- Verify — six hours later the pair is read again and the forecast is marked RUG, FADED or HELD against what actually happened.
- Revise — instruments that called it earn weight, instruments that cried wolf lose it. Each revision is a new model run, logged with its sample size and reasoning.
03The bands
Three, and they mean what they sound like:
04The twelve instruments
Each is a plain rule with a weight and hard bounds that verification may never push it past. Every instrument that fires on a reading is listed on that station's report, with what it was worth at the time — so any forecast can be re-derived by hand.
→ see what each is currently worth
05Missing data is never a zero
Sources omit fields. DexScreener leaves liquidity out entirely on very fresh pairs; the RPC occasionally refuses. When that happens the instrument does not fire and is reported as unread — it is never quietly treated as a reading of zero, which would turn silence into evidence. Every station report shows its coverage, and the chart counts stations it could not place rather than placing them at the origin.
06Where the numbers come from
| source | endpoint | supplies |
|---|---|---|
| DexScreener | api.dexscreener.com | pair discovery, liquidity, volume, age, price move, order flow |
| Solana RPC | api.mainnet-beta.solana.com | mint authority, freeze authority, wallet holdings |
Both are public and neither needs a key, so there is nothing secret in this deployment. If a source stops answering, the affected readings show as unread and the bulletin says the office is not reporting.
07What this is not
It is not advice, and it does not trade. A CLEAR reading means twelve instruments found little wrong at one moment — not that a pair will go up, and not that it is safe to buy. Those are different claims and the office only makes the first. Nothing here places an order, holds a key, or asks for a signature.
08Judging the office
Do not take the forecasts on faith. The verification record shows how often each band was wrong, and the model page shows every weight revision with the sample that caused it. If the CLEAR band stops beating the field, that will appear there before it appears anywhere else.