BRENNEN

Industries / Mining and materials / Chemicals and petrochemicals

Specialty chemicals

The specialty chemicals industry makes small volumes of chemicals that are sold for what they do, such as coatings, adhesives, additives, catalysts and flavors, rather than for what they are.

1874

Chemists Wilhelm Haarmann and Ferdinand Tiemann made vanillin, the main flavor compound of vanilla, from a compound in the sap of conifer trees and started a small factory for it in Holzminden, Germany. A history of synthetic vanillin dates the start of its industrial-scale production to that factory, later Haarmann & Reimer and today part of Symrise.

sabersenaccio.iec.cat

35

Milestones achieved across 6 domains.

12

To watch for: announced, not yet confirmed outside.

12

Open deliverables. Nobody has shipped them.

In one minute

Where it started

Specialty chemicals are bought for an effect, such as a taste, a color, a seal or a faster reaction. The trade began in 1874, when Haarmann and Tiemann started making vanillin in a factory, a flavor that until then came only from vanilla beans. Paints thinned with water, glues that set in seconds and catalysts sold to oil refiners followed in the twentieth century, most of them made a few tonnes at a time in batch vessels.

Where it stands

Suppliers now sell AI tools alongside their products: in 2026 3M launched an assistant that recommends adhesives and Henkel one that simulates glued joints, and neither has published how often it is right. Large groups are reshaping their businesses: in June 2026 BASF completed the sale of most of its coatings business to the investment firm Carlyle, at an enterprise value of €7.7 billion. In Europe, a committee of the European Chemicals Agency backed restricting PFAS as a whole class of chemicals in March 2026, and the final decision is still pending.

What is open

Much of what the industry knows is scattered: recipe tables sit in patents, reaction heats in papers, lab and outdoor corrosion results in separate studies, and hazard data on millions of safety data sheets. Many of the open jobs below turn that material into checked public tables, using AI models that read long documents. Others put numbers on claims that nobody has tested in public, such as how often AI glue advisers or in-store paint color matches are right.

Domains

AchievedWatch forOpen

Advances to build with

5 research notes

Open in specialty chemicals

3 of 12 open deliverables

Open · Formulation

Recipe tables from patents

A free, checked dataset of the worked examples in adhesive and coating patents. Each row gives one example's ingredients, their amounts, how it was cured, and the test results the patent reports, such as lap-shear strength (the force needed to slide two glued strips apart) or pencil hardness. Add a simple model that predicts a test result from the recipe, scored on patents it has not seen.

Built with what exists

Google Patents full text, searched by patent class (C09J for adhesives, C09D for coating compositions such as paints); a small AI model built for extraction and long documents, such as Meta's Muse Ember (see the Note "The jobs agents repeat"); RDKit, an open-source chemistry toolkit, to link ingredient names to structures where a patent gives a chemical name; CheMixHub's code for data splits and baseline models.

Credit · unclaimed

Open · Batch production

Procedures into checked batch recipes

Software that reads a written batch procedure, such as a worked example in a patent, and writes it as a structured recipe in BatchML, with each charge, heating step, hold and transfer as its own step. A second program checks the recipe against plain rules a reviewer applies: a reactive ingredient is fed in portions or at a set rate rather than all at once, every heating step has a maximum temperature and a cooling action, every hold has a time. It then lists what the written text leaves out.

Built with what exists

The BatchML schemas, free from MESA International on GitHub; RWTH Aachen's open-source RecipeEditor, which draws and validates BatchML recipes; IBM's open-source Paragraph2Actions as a baseline for splitting text into steps; worked examples from patents, which are free to read; an open-weight model to read the text, such as Mistral Large 4 (see the Note "A trillion open parameters"); a coding agent that tests its own output, such as one built on Inkling 2 Small (see the Note "Checked without being asked"). ISA forbids entering its standards into AI tools, so the rules should be written from the public BatchML schema and published guidance, not by pasting the standard.

Credit · unclaimed

Open · Scaling from lab to plant

Open table of reaction heats

A public table of measured heats for the main reaction in published process-development studies: heat of reaction, heat capacity of the mixture, adiabatic temperature rise (how hot the batch would get with no cooling at all), the highest temperature reached if cooling fails, the onset of decomposition, the instrument used, and a link to the page. Then a test of how closely free estimating methods predict the measured heats.

Built with what exists

Open-access process papers, such as the free CHIMIA archive; a model that reads long documents and tables, such as Muse Ember (see the Note "The jobs agents repeat"); the Open Reaction Database's open schema, extended with heat fields; thermochemistry estimates from the open-source Reaction Mechanism Generator (RMG), maintained at MIT and Northeastern University.

Credit · unclaimed

Where it sits