Summary
The European Patent Office (EPO) announced the winners of CodeFest 2026 at the PATLIB conference in Warsaw on 16 September 2026. The competition awarded AI-driven platforms designed to evaluate patent portfolio value by assessing legal strength, commercial blocking power, and substantive inventive step beyond traditional citation counts. For IP counsel and portfolio managers, these developments point to new methodologies for automated portfolio audits, valuation for financing, and prior art screening during prosecution.
The Event
During the annual PATLIB conference in Warsaw, Poland, the EPO presented awards for its fourth CodeFest competition, which drew 23 submissions from 12 countries. The jury of EPO and external data science experts selected six finalists and three winning teams:
- Grand Prize (EUR 20 000): Team Confused Electrons (Arnab Saha, Ralph Ryan Hebrio, Mauricio Rodriguez Alas). The team built a machine learning platform that evaluates long-term strategic value across patent portfolios by linking technological impact, market relevance, and legal strength to market trends.
- First Runner-up (EUR 10 000): Team Red-Cube-Coders (George Abi Younes, Mehdi El Bouari, Reza Hosseini, Elena Mas Tur, Gaétan de Rassenfosse). Their tool quantifies technical and market value using two metrics: a "blocking score" to measure constraints imposed on competitors, and a "commercialisation score" to measure market potential.
- Second Runner-up (EUR 5 000): Team EUREKA (Grazia Sveva Ascione, Nicolo Tamagnone). The solution models patent examiner reasoning, identifying closest prior art and analyzing technical differences to assess inventive step rather than relying on simple keyword matches.
Prizes were presented by Ewa Skrzydło-Tefelska (President of the Patent Office of the Republic of Poland), Silvia Rossi (Director of the State Office for Patents and Trademarks of San Marino), and Muhammed Zeki Durak (President of TÜRKPATENT), with the ceremony moderated by EPO Chief Information Officer Diego Eguidazu Alonso.
Context
Traditional algorithmic IP valuation relied heavily on bibliometric indicators, such as forward citation counts, family size, and renewal maintenance. While easy to aggregate, these metrics lag real-time prosecution and market reality. Citation data reflects past activity and can fail to capture whether a patent presents an enforceable technical hurdle.
The 2026 competition focused directly on bridging this gap. Finalists demonstrated models that integrate qualitative patent examination concepts, semantic analysis, and economic indicators. Alongside tools such as the EPO Deep Tech Finder, Data Desk, and the modernized IPscore framework discussed at the PATLIB conference, the office is encouraging methods that combine legal analysis with automated machine learning.
Implications
The emergence of AI tools modeling examiner reasoning and competitive blocking has concrete operational consequences for in-house IP leaders and prosecution attorneys:
- Substantive Prior Art Audits: Systems modeling examiner-style reasoning, like EUREKA, demonstrate that natural language processing can parse technical differences against closest prior art. Practitioners can deploy similar internal tools prior to filing at the EPO or KIPO to test claim distinguishability and anticipate inventive step rejections.
- Pruning and Maintenance Decisions: Rather than relying solely on business unit sentiment or raw citation counts, portfolio managers can evaluate commercialization and blocking scores to identify non-core assets for abandonment or licensing.
- Relevance for Asian and Korean Filers: Korean applicants holding large European portfolios often face high maintenance and validation fees across EPO member states. Utilizing structured blocking metrics allows Korean corporate IP teams to objectively defend budget allocations and identify which European filings truly restrict competitors versus those offering minimal technical coverage. In Korea, while KIPO practice features strict amendment requirements under Article 47(2) of the Patent Act, pre-filing algorithmic checks against closest prior art help align European and Korean claim drafting before entering regional phases.
Action for practitioners: Audit your current portfolio assessment criteria. If your organization relies purely on forward citations or renewal age to rank asset importance, incorporate qualitative indicators such as claim scope breadth, freedom-to-operate blocking potential, and technical distance from primary prior art into your periodic maintenance reviews.
Outlook
The integration of AI-assisted valuation into mainstream EPO tools, including the ongoing development of the new IPscore project, indicates that patent offices and institutional investors will increasingly evaluate patent quality using automated technical benchmarking.
While these tools continue to mature, IP owners should review the technical disclosures in their pending priority applications. Ensure that specifications explicitly describe technical effects and distinguish claimed features from known solutions, providing clear support to withstand both examiner reasoning models and automated prior art comparisons during examination.