The salting master knows which barrel is ready — but she cannot taste them all today.
Curing Cantabrian PDO anchovy means cross-referencing the salinity of the brine, the temperature of the chamber and the time in salt for every barrel — data that lives in a notebook, in a datalogger and in a spreadsheet. iLEAN Brain cross-references the three sources continuously, ranks which barrels are entering their window and proposes the sampling order to the master. She tastes, decides and signs off.
Hundreds of barrels, three data sources that never talk to each other, one tasting window.
Curing Cantabrian PDO anchovy is a long process in which the final decision depends on cross-referencing three realities that, in most salting houses, each live on their own island:
- Brine salinity per barrel — the operator measures it with a refractometer and writes it in a notebook. Sometimes twice a week, sometimes once, depending on the barrel.
- Chamber temperature — the supplier's datalogger or the analog thermometer. Rarely integrated with anything else.
- Time in salt per barrel — on the intake sheet the master keeps on paper or in her own spreadsheet.
The master knows which barrel is ready when she tastes it — but she cannot taste all 1,200 at once, and the optimal window lasts only a few days per barrel. Open it late and the over-matured anchovy drops a PDO grade; open it early and the tasting comes out salty and tough. The craft works. What the craft cannot do is be in 1,200 places at once.
iLEAN Brain — the salting master's notebook, alive, cross-referenced and prioritized.
The problem with curing is not a lack of judgment in the master: it is cross-referencing three heterogeneous sources sustained over time, across hundreds of barrels. iLEAN acts as the putty between the salinity notebook, the chamber datalogger and the salt-in sheet, without replacing the tasting decision.
Connect captures the three sources — whether the data comes from the refractometer, the datalogger or the notebook. Brain cross-references, learns and ranks. The master tastes and signs off. The system opens nothing.
The three iLEAN pieces applied to Cantabrian PDO anchovy curing:
- iLEAN Connect — captures through whichever door is open. A photo of the brine notebook at the end of the round. A camera reading of the old datalogger's display. Direct integration with a modern datalogger if there is one. The operator no longer fills in a spreadsheet at night; the data enters at second zero.
- iLEAN Brain (Agents) — they live on Central and continuously cross-reference the three sources with the history of previous tastings. They learn which barrels ended up with a premium tasting and what pattern they followed, and they propose a sampling order: "barrel 142 has been in its optimal window for 3 days, top priority; 167 is running late because of temperature, take an intermediate sample".
- Edge (optional) — if the salting house is digitizing its sampling, an analytical balance or a colorimeter can be integrated through Edge to enrich the tasting data.
A salting house with a notebook and a spreadsheet vs. a salting house with iLEAN Brain
| Aspect | Classic salting house | With iLEAN Connect + Brain |
|---|---|---|
| Salinity per barrel | Notebook, transcribed by hand | Digitized at second zero |
| Chamber temperature | Isolated datalogger | Continuously cross-referenced with every barrel |
| Tasting priority | Memory and visual rotation | Ranked list of barrels in their window |
| The master's craft | In her head, and it leaves with her | Captured as a pattern, tasting by tasting |
| Risk of opening a barrel late | High over the last days of the window | Flagged days in advance |
| Per-barrel dossier for the PDO | Rebuilt by hand, over weeks | Automatic, with every reading included |
Impact estimate for your plant — to be validated with your numbers.
The block below is an estimate to be validated with the specific data of your salting house. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.
- Mid-sized PDO salting house, several hundred to a little over a thousand barrels curing at once, an annual season, a team with a salting master and operators.
- Connect pilot to digitize the brine notebook + integration of the chamber datalogger + Brain with the tasting history. First value expected within a few weeks (the first genuinely useful ranked sampling list).
- Indicative payback between 4 and 9 months, depending on volume and on the documented loss from barrels opened outside their window.
- Reduction in barrels tasted outside their window ≥30% (a conservative estimate). The hard lever is threefold: less loss of PDO grade, a better average tasting in the final batch, and the master's craft captured as a pattern before she retires.
And the master's reasonable doubt
"What if the AI gets it wrong and tells me to taste a barrel that isn't ready?" — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI cross-references real data from the system itself (salinity, temperature, days in salt) against a learned pattern, the best models brought error below 1.5% [1]. And even so, what is critical is never decided alone: iLEAN proposes the sampling order, the master tastes and signs off. The system does not open barrels — the master does.
[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.
What people ask about Cantabrian PDO anchovy curing with agents
What makes Cantabrian PDO anchovy curing unique?
Curing Cantabrian anchovy under a Protected Designation of Origin is a process that runs for months (typically 6 to 12, varying barrel by barrel) during which fresh anchovy landed in the season is salted in barrels with sea salt, pressed under weight and matured in a temperature-controlled chamber. The optimal point is set by the combination of the salinity of the brine the fish itself releases, the sustained temperature of the chamber and the time in salt — and the salting master reads it from experience. Open the barrel too early and you get a salty, tough anchovy; open it too late and you get an over-matured anchovy that loses its aroma. The PDO rules set the organoleptic grades; the craft sets the exact day.
Why is anchovy curing a textbook problem for AI agents?
Because it is a long process, with many barrels running at once, with data living in different places (manual salinity readings in a notebook, chamber temperature in the datalogger, the salt-in date on the master's sheet), and where the final decision blends human judgment with the cross-referencing of three variables. iLEAN Brain (the Agents running on Central) is exactly the piece for this: cross-referencing heterogeneous sources continuously, anticipating which barrel enters its window ahead of the rest, and proposing the tasting order to the master — who signs off.
How does the data get in if the brine readings are taken by hand?
iLEAN Connect captures through whichever door is open. If the operator measures salinity with a refractometer and writes it in a notebook, a photo at the end of the round is enough — Connect digitizes the notebook. If the chamber temperature sits in an old datalogger, Edge reads the display with a camera. If there is a modern Modbus sensor, it integrates directly. Capture is graduated machine by machine, not project by project — we capture wherever data is available. The master no longer fills in a spreadsheet at night; the Agents do the cross-referencing at second zero.
What exactly does the agent do with that data?
The Agents (iLEAN Brain) live on Central and work 24/7. They cross-reference the salinity of each barrel with the chamber's temperature curve and the days since it went into salt. From the salting house's own history they learn how the barrels that ended up with a premium tasting behaved compared with those that ended up in a lower grade, and they propose a prioritized sampling list to the master: "barrel 142 has been in its optimal window for 3 days, worth tasting; 167 is running late because of temperature, I suggest an intermediate sample". The master tastes, decides and signs off — the system opens nothing.
What does an agent pilot on an anchovy salting house cost?
A Brain/Agents pilot on a mid-sized salting house comes at a reasonable order of magnitude: a Connect terminal to capture the sources (brine notebook, chamber datalogger, salt-in sheet), the Central licence with the Agents, and the tuning against the master's own history. The indicative payback (an estimate to be validated) is between 4 and 9 months, depending on the salting house's PDO volume and the documented loss per barrel opened outside its window. The hard lever is threefold: fewer barrels outside their optimum, less risk of dropping a PDO grade, and the master's craft captured as a pattern you can reuse every season.
Tell us your case and in 48h we'll send you the estimated ROI of this AI project for your PDO anchovy salting house.
We work on the real data from your salting house, not ours. Diagnostic with no commitment.
Request estimated ROI in 48h See the food industry