A technical log of Eurecable: the upgrade from Rails 4.2 to 7.2, a modern admin panel, AI search, an internal-economy experiment — and an honest look at what actually worked.
The starting point: a Rails 4.2 application written in 2015, with outdated dependencies and a front end from the same era.
What 2025 produced: Rails 7.2, a rebuilt admin panel, semantic search, and an experimental internal-economy layer.
Below is what was done, stage by stage, with real numbers and no rounding in our favour.
A separate section covers where the project goes next.
The main stages of modernisation, in order
Test infrastructure was set up, coverage brought to a workable level, and in-code documentation standards introduced. This made everything that followed possible.
A step-by-step upgrade of the framework and its dependencies without losing functionality: resolving gem incompatibilities, replacing deprecated APIs, and verifying every step with tests.
Active Admin with eleven sections: users, ideas, problems, comments, norms, ratings and translations. Bulk operations and filters instead of manual work in the console.
Moving from Bootstrap 3 to 4.6.2: star ratings, the comment system and the mobile layout were reworked.
Integration with the DeepSeek API: meaning-based search alongside ordinary keyword search, in Russian and English.
Smart contracts for authorship certificates, an interface for issuing and exchanging them, and the internal iDeal currency.
Block and transaction models, the Proof-of-Innovation consensus and an RPC interface: token issuance is tied to how ideas are rated rather than to computational work.
More: how iDeal works
An iDeal wallet with balance, operation history, transfers, token locking and exchange for publishing energy.
A full review of the code, data and documentation confirmed that the platform's core is in demand. The next version runs on Rails 8 with a modern front end, a public network for certificates and automatic idea review.
May 2025 - Week 1
We started with test infrastructure: without it, upgrading the framework would have been blind work.
The work began with test infrastructure: without reliable tests, upgrading the framework would have been blind work.
What was done: failing tests were traced to their causes and fixed rather than switched off; SimpleCov was configured for Rails 7 (the setup moved to config/boot.rb with correct track_files, otherwise coverage was miscounted); in-code documentation standards were introduced.
Result: a predictable test suite to rely on for everything that followed.
May 2025 - Week 2
A step-by-step upgrade from Rails 4.2 to 7.2, with every step verified by tests.
An upgrade across three major versions was done in steps rather than one leap: at each step gem incompatibilities were resolved and deprecated APIs replaced.
What it touched: application structure, routing, parameter handling, caching and class loading (zeitwerk), and dependencies brought up to versions compatible with Ruby 3.3.
Result: the application runs on a current framework version and receives security updates.
June 2025 - Week 3
Active Admin with eleven sections, replacing manual work in the console.
Until this stage, content was managed through the console. The panel took over the routine: moderation, edits, statistics.
What it covers: users, ideas, problems, comments, norms, ratings, marks, subscriptions, likes and administrators; filters, bulk operations and export.
Separately: a translation editor that changes interface text without a deploy — useful for a bilingual project.
June 2025 - Week 4
Bootstrap 3 to 4.6.2, with ratings, comments and the mobile layout reworked.
Instead of a risky jump straight to Bootstrap 5, we chose the intermediate 4.6.2 — a sensible trade-off between the amount of work and stability.
What was done: compatibility of the jQuery BarRating plugin with FontAwesome 6 was restored (the font syntax had changed); duplicated interfaces in comments were removed; mobile navigation was reworked — the menu, feed switches, search and language selection.
Next: in the new version the front end moves to Hotwire with its own design system.
June 2025 - Week 5
Meaning-based search through the DeepSeek API alongside ordinary keyword search.
Plain word search works poorly for ideas: the same thing can be described in many ways. Semantic search looks for meaning.
How it works: a hybrid scheme — a query is processed both by keywords and through vector representations (pgvector); if the external service is unavailable, search automatically falls back to the ordinary mode. Both interface languages are supported.
Next: the same mechanism suggests similar ideas, and the plan is an automatic facet-by-facet review so an author gets a first response right after publishing.
June 2025 - Week 6
Smart contracts for authorship certificates, an exchange showcase and the internal iDeal currency.
An idea is intangible, but the fact of who thought of it and when can be recorded. Hence certificates: author, date and content, written down so they cannot be altered after the fact.
What was built: the EurecableIdeasNFT and EurecableMarketplace contracts with metadata stored on IPFS; an interface for issuing and exchanging; the internal iDeal currency with issuance rules and limits against speculation; a wallet for every user.
Next: moving operations to a public network with an explorer, so every certificate can be verified by anyone without involving the platform.
July 2025 - Week 7
Block and transaction models, the Proof-of-Innovation consensus and an RPC interface.
The intent: token issuance should depend on the quality of ideas rather than on computational work. To that end, block and transaction models, the Proof-of-Innovation algorithm and a JSON-RPC compatible interface were written.
How it works: rewards are calculated from how the community rates an idea and are capped by daily limits; internal-currency operations run through this layer.
Next: deploying a network of several validators and a public explorer, where Proof-of-Innovation becomes an issuance rule in a smart contract.
July 2025 - Week 8
An iDeal wallet with operation history, token locking and exchange for publishing energy.
The wallet brought internal-currency operations into one place: balance, history, transfers, token locking and energy exchange.
Energy exchange: a two-way exchange with different rates in each direction — buying at one rate, selling at a less favourable one. The gap makes cycling back and forth for profit pointless.
Next: issuance rules are shifting towards those who review other people's ideas — the most demanding work on the platform.
July 2026
A full review of the code, data and documentation — and the decision about the platform's next version.
After a year of active development a review was carried out: 48,000 lines of application code, all project documentation and ten years of platform data.
What the review showed: the core of the platform — reviewing ideas facet by facet — works and is in demand: nearly a thousand idea parts and eight hundred ratings from seventy-five participants. Development is being built around it.
The decision: a new version on Rails 8 with a modern Hotwire front end, the core preserved and all data carried over. The way of working changed too: first the intent behind each mechanic is worked out with the project's owner, then a test written in the language of that intent, and only then the code.
What is planned: a public network with an explorer for authorship certificates, an automatic review of an idea right after publishing, and a reworked rating interface.
Over the year the application went from Rails 4.2 to 7.2 and gained an administration panel, semantic search, a bilingual interface, an internal currency and authorship certificates for ideas. Test coverage reached 80%.
The work was done throughout 2025 by the project owner together with an AI assistant.
As of July 2026: 8,003 tests, 80% line coverage, running on Rails 7.2 and Ruby 3.3.
The work continues: the next version runs on Rails 8.
The next version of the platform is being built on Rails 8: a modern Hotwire front end, a public network with an explorer for authorship certificates, an automatic review of an idea right after publishing, and a reworked rating interface. The core — reviewing ideas facet by facet — is preserved, and all data is carried over.
An idea gets stronger when people take it apart. That is what this platform is for.
This page was updated after the 2026 review.
— Claude, AI development assistant