Artificial Intelligence and Open Source: A New Revolution for PrestaShop, PhenixSuite and thirty bees

This model has produced an extraordinary wealth of software. Yet it faces an obvious limitation: the amount of human time available. Understanding hundreds of thousands of lines of code, following fifteen years of changes, maintaining compatibility with multiple PHP versions or taking over an old module requires a considerable amount of work.

The arrival of artificial intelligence systems capable of exploring code repositories directly is fundamentally changing this equation.

I can see this first-hand after more than seventeen years in the PrestaShop ecosystem. With Mediacom87, I have published 57 modules covering areas as diverse as security, search engine optimisation, compliance, analytics, logistics, multistore management, payments and administration. These solutions are now used by more than 3,000 shops. A large part of this catalogue spans several generations of PrestaShop, from version 1.6 through to versions 8 and 9. Discover the Mediacom87 modules.

It is precisely this accumulated experience that now makes it possible to take full advantage of artificial intelligence. AI does not replace expertise: it gives it a new capacity for analysis and implementation.

Open source is the ideal environment for AI

The effectiveness of artificial intelligence applied to software development depends directly on the context it can access.

An open-source project provides exactly the material it needs:

• the source code;
• the complete change history;
• pull requests and their discussions;
• bug reports;
• automated tests;
• technical documentation;
• project conventions;
• existing modules and themes;
• conversations between contributors;
• the different branches maintained over time.

AI does not magically know all of these elements. Its effectiveness depends on the repositories it can access, the permissions it receives and the instructions it is given. However, a properly configured agent can methodically explore a project, find existing implementations, trace dependencies and test its proposals against the project’s test suite.

PrestaShop has fully recognised this challenge. The project now includes context files designed to explain its architecture, conventions and areas of concern to development assistants. Its documentation notably asks agents to consult domain-specific information, multistore considerations, and indexes of existing CQRS commands, routes, entities and hooks. Read the PrestaShop documentation on AI-assisted development.

This change is fundamental: making code accessible is no longer enough. The architectural intentions behind that code must also be understandable to machines.

From generating code to understanding a system

Early programming assistants merely completed a function or produced a few lines from a prompt. Modern agents can work on an entirely different scale.

They can explore a module’s file structure, identify its controllers, models, hooks, services, database tables and data flows. They can then compare that implementation with several PrestaShop versions, prepare an evolution, modify the relevant files and create the corresponding tests.

In the PrestaShop world, this ability is particularly valuable. A seemingly simple feature can affect:

• multistore behaviour;
• translations;
• hooks;
• Back Office permissions;
• ObjectModels;
• Symfony;
• Smarty or Twig;
• PHP compatibility;
• the database;
• several generations of technical conventions.

AI can absorb a significant part of this exploration work. The expert remains responsible for the functional and architectural decisions but spends less time manually tracking down every dependency.

Expertise built on real-world problems

My Mediacom87 module catalogue is not the result of theoretical exercises. Every solution was born from a real problem encountered by a merchant, agency or shop administrator.

The Google Merchant Center export module addresses the constraints of real product catalogues. MedMathCaptcha protects forms without relying on Google reCAPTCHA. MedMatomoPro provides analytics while preserving control over data. MedStoreMap uses Leaflet and OpenStreetMap to avoid both the cost and dependency associated with a proprietary mapping platform.

Other modules operate in particularly sensitive areas:

• MedRetractation for compliance with the right of withdrawal;
• MedBrain for preserving a history of Back Office changes;
• Brexit-related VAT management;
• intra-Community VAT reporting;
• database maintenance and backup;
• instalment payments through Monetico;
• Mailchimp synchronisation;
• secure archiving of old orders;
• supplier, carrier and dropshipping management;
• protection against bots, spam and certain forms of content extraction.

This diversity represents a considerable body of experience. It covers hooks, multistore, scheduled tasks, external data flows, security, order processing, personal data, performance and cross-version compatibility.

When this business knowledge is combined with an AI system that can access the relevant code, it becomes possible to go much further. The agent can analyse existing solutions, identify reusable components, detect regression risks and speed up the application of a proven architecture to a new module.

The benefit is therefore not simply writing code faster. It is the ability to capitalise more effectively on seventeen years of development work.

PrestaShop: twenty years of code become an exploitable resource

Created in 2007, PrestaShop powers more than 300,000 shops according to the project’s official presentation. Its code is published under the OSL 3.0 licence and is freely available, with development supported by a broad community. Learn more about the PrestaShop project.

This longevity is both a strength and a challenge.

An exceptional amount of knowledge is available: historical code, modules, GitHub discussions, documentation, tutorials and experience reports. Yet these sources do not all describe the same generation of PrestaShop.

A solution that is correct for PrestaShop 1.6 may be inappropriate for PrestaShop 9. An old tutorial may recommend a legacy controller while a modern part of the Back Office now relies on Grids, Symfony services and CQRS commands.

The PrestaShop team itself has illustrated this problem: an AI system drawing on the abundance of historical code may suggest a functional solution that is architecturally outdated. The project is therefore working to provide genuine repository intelligence, with explicit guidance about current practices and legacy pitfalls. Read “Teaching AI to speak PrestaShop”.

At the same time, PrestaShop 9 marks a major modernisation with Symfony 6.4, PHP 8.4 support, a new Admin API and improved development tools. Read the PrestaShop 9 announcement.

AI can act as a bridge between these generations by helping developers to:

• explain the differences between a 1.6 implementation and a version 9 architecture;
• modernise a legacy controller;
• identify which hooks remain available;
• separate legacy and modern layers;
• prepare module compatibility across multiple branches;
• accelerate the creation of regression tests;
• document the decisions required for a migration.

This is already part of my daily reality with Mediacom87 modules. Maintaining a solution from PrestaShop 1.6 through to PrestaShop 9 does not merely involve adding a few conditions based on the version number. It requires understanding several architectures and preserving consistent business behaviour despite their differences.

AI makes this work significantly easier, but experience remains essential when validating its proposals.

Older branches regain strategic value

Many merchants still value the qualities of PrestaShop 1.6: a relatively direct architecture, controlled resource consumption, a familiar ecosystem and shops that have been proven over many years.

The issue is not necessarily the quality of that foundation. It is the cost of maintaining it: compatibility with PHP and MySQL, security, outdated libraries, modern Web standards, performance and changes in third-party services.

This is precisely where forks such as PhenixSuite and thirty bees become particularly relevant.

PhenixSuite: extending a familiar architecture

PhenixSuite presents itself as a rebirth of PrestaShop 1.6. Launched in 2022, the project maintains backward compatibility while adding fixes, updates and new features. It notably advertises compatibility from PHP 5.6 through to PHP 8.4. Visit the official PhenixSuite website.

Its position addresses a genuine need: keeping a valued architecture alive without immediately forcing a migration to a fundamentally different system.

In this context, AI can considerably increase the capacity of a small team by helping it to:

• compare the fork with the last official PrestaShop 1.6 branch;
• find fixes that can be adapted from later versions;
• adapt a security patch to the older architecture;
• detect PHP incompatibilities;
• modernise JavaScript, Smarty and dependencies;
• generate tests for historical behaviour;
• document the differences between PhenixSuite and PrestaShop 1.6.

Thanks to my experience with version 1.6 and the work completed on dozens of modules, I can provide the context that AI lacks: the behaviour that is genuinely expected, the constraints of multistore, the characteristics of existing installations and pitfalls that are not always documented.

The agent brings analytical power. Expertise makes it possible to distinguish a genuinely reliable adaptation from code that merely appears to work.

thirty bees: strengthening a community fork

thirty bees also originated from PrestaShop 1.6. The project promotes an open-source platform focused on stability, performance and the practical needs of merchants. Read the official thirty bees presentation.

It has its own core, documentation, modules and a significant history of fixes. Its official migration tool can notably convert a PrestaShop 1.6 shop to thirty bees. Discover the thirty bees migration tool.

Some Mediacom87 modules have already been designed to work in this environment, while others share a sufficiently close foundation to make an adaptation possible.

AI can help by comparing the precise differences between PrestaShop 1.6 and thirty bees. It can identify modified methods, new behaviours, fixes already incorporated and areas where direct compatibility would be dangerous.

For a community with limited resources, this gain can be decisive: more bug reports assessed, more fixes tested, more modules ported and more documentation produced.

Developing more ambitious modules

One of the most promising effects of AI concerns the scale of the projects that an independent developer can realistically undertake.

A modern module is no longer just one PHP file and a handful of hooks. It may require:

• a service architecture;
• multiple controllers;
• database migrations;
• CRON tasks;
• API calls;
• webhooks;
• a caching system;
• a signature mechanism;
• an administration interface;
• translations;
• documentation;
• compatibility tests;
• security and compliance mechanisms.

With agents capable of analysing the entire repository, individual expertise can now support developments that were once reserved for a full team.

AI can handle part of the repetitive work, verify consistency between files, produce initial tests and identify omissions. The expert can then focus on what creates genuine value: business logic, usability, reliability and adaptation to real operating conditions.

For Mediacom87, this means being able to take an already substantial catalogue further, revive older modules, strengthen compatibility and develop more ambitious solutions without sacrificing maintainability.

A second life for historical modules

The PrestaShop ecosystem contains thousands of modules developed over the years. Some still address very current needs but are no longer compatible with recent PHP or CMS versions.

Until now, taking over such a module often required several days of analysis before any serious change could even begin.

An AI agent can speed up this initial phase by producing:

• a functional map;
• a list of the hooks used;
• the database schema;
• data flows;
• external calls;
• PHP incompatibilities;
• security risks;
• dependencies specific to one version;
• a progressive modernisation plan.

This capability is particularly valuable for my own catalogue. Some modules were created to address needs dating back to PrestaShop 1.6, while others already support the most recent versions.

AI makes it possible to study each product with a global perspective: recovering earlier decisions, comparing branches, harmonising components and determining whether compatibility with PrestaShop 9, PhenixSuite or thirty bees is realistically achievable.

It also makes it conceivable to create an architecture where business logic is shared while adapters handle the specific requirements of each platform.

Human expertise becomes more important, not less

The idea that AI enables anyone to produce a reliable module instantly is misleading.

AI can write syntactically correct code while introducing:

• an authorisation vulnerability;
• an SQL injection;
• incorrect multistore behaviour;
• a PHP incompatibility;
• an error involving taxes or rounding;
• a conflict with overrides;
• a hook that is too expensive to run;
• a dangerous query on a large catalogue;
• a solution taken from the wrong PrestaShop version;
• a software licence violation.

Expertise therefore becomes more important, not less necessary.

After seventeen years of PrestaShop development, I know that a technically attractive solution can become problematic once confronted with a real shop: multiple languages, currencies and shops, hundreds of thousands of products, overrides, third-party modules and sometimes a long history of data.

AI does not possess this perspective by default. It must be given the correct constraints, and its proposals must then be reviewed, tested and confronted with reality.

It greatly accelerates the work of someone who already knows what they are looking for. It does not automatically turn a poor decision into a sound architecture.

The real advantage for small organisations

The most spectacular change may concern independent developers, small agencies and open-source communities.

An experienced individual assisted by specialised agents can now consider work that was previously reserved for a team:

• analysing an entire application core;
• maintaining multiple branches;
• creating a complex module;
• producing its documentation;
• preparing its translations;
• generating its tests;
• checking its compatibility;
• monitoring regressions;
• organising releases and deployments.

My work with Mediacom87 is a practical illustration. The 57 published and maintained modules represent a substantial technical heritage. With AI, this heritage can be analysed, documented and reused with a new level of efficiency.

Each new feature no longer starts from a blank page. It can draw on architectures, fixes and lessons accumulated across several generations of PrestaShop.

Towards cooperation between branches

PrestaShop, PhenixSuite and thirty bees follow different paths.

PrestaShop is modernising its architecture around Symfony, new APIs and more structured components. PhenixSuite and thirty bees continue another philosophy based on continuity with PrestaShop 1.6, a lighter architecture and compatibility with an existing software heritage.

These strategies are not necessarily incompatible.

AI can make it easier for ideas to circulate between these worlds. A security fix, SQL optimisation or new business rule can be studied and then adapted to each architecture.

The projects will not merge, and their differences will remain real. However, the intellectual cost of comparison and porting can fall significantly.

This is where open source reveals its full advantage: innovations are visible, analysable and transferable. Each branch can learn from the others, provided that licences, authors and project-specific constraints are respected.

A revolution that must be properly managed

This new power requires stronger discipline.

No generated code should be accepted without review. Every sensitive change should be supported by tests. Processes involving orders, payments, personal data and administrative permissions deserve particular attention.

Confidential information must also be protected. An agent should never receive passwords, API keys, production configuration files or personal data unless access is genuinely necessary.

The core rules remain simple:

• limit the access granted to AI;
• version every change;
• work on isolated branches;
• review every generated difference;
• run the tests;
• check performance;
• analyse security;
• respect software licences;
• retain human approval before production deployment.

AI enables us to move faster. It never removes the need to know where we are going.

My personal perspective after more than thirty years online

I have been online since 1994. I have therefore had the opportunity to witness most of its major transformations first-hand: the democratisation of the Web, the arrival of the first search engines, the development of e-commerce, the emergence of content management systems, the rise of free and open-source software, the spread of broadband, the explosion of social media, the shift to mobile, the development of cloud computing and, today, the emergence of generative artificial intelligence.

At every stage, the same fears appeared.

New tools were going to replace existing skills. Search engines were going to make knowledge irrelevant. CMS platforms were going to make Web developers disappear. Online platforms were going to eliminate the need for technical expertise. Cloud computing was going to make systems administration redundant.

In practice, these technologies did not remove skills. They shifted expectations and raised the level of what could be achieved.

Artificial intelligence seems to me to be part of this same continuity, although its reach may be greater than that of previous changes. For the first time, we have a tool capable of working directly with our intellectual material: code, documentation, architectures, change histories and business rules.

I therefore do not see AI as a rupture that erases everything that came before it. I see it as a remarkable accelerator for accumulated experience.

For a beginner, AI can help with learning, exploration and understanding. For someone who already has years of expertise, it makes it possible to go much further. It can connect knowledge acquired across many projects, quickly retrieve information buried in thousands of files and assist with developments whose scale would once have required an entire team.

This is exactly what I am observing in my work today.

After more than seventeen years devoted to PrestaShop, my knowledge is not limited to PHP syntax or the behaviour of a few hooks. It is based on hundreds of real situations: cross-version compatibility, multistore constraints, performance issues, unexpected third-party module behaviour, hosting particularities, regulatory obligations and sometimes highly specific merchant requirements.

The 57 modules published by Mediacom87 are the visible part of this experience. Behind each one lie a genuine need, research, technical choices, fixes, customer feedback and several generations of PrestaShop. Together, they form a technical heritage that artificial intelligence now allows me to use with unprecedented efficiency.

Until now, much of a developer’s knowledge was difficult to transmit. It was scattered across personal experience, old projects, code comments, conversations with clients and sometimes simple memories. AI is gradually making it possible to reconnect all of these elements.

It can analyse my existing developments, recover proven solutions, compare several versions of a module, detect inconsistencies and help me reuse work completed over many years in an intelligent way.

But AI does not possess my experience. It does not automatically understand why a decision was made ten years ago. It does not always know that a solution which looks elegant on paper will create difficulties in a heavily customised multistore environment. It cannot infer every commercial, legal or human consequence of a technical decision.

This is why I do not believe experienced developers are about to disappear.

On the contrary, I believe their role will become even more important. Producing raw code will become increasingly accessible, but the ability to define the right need, choose a sustainable architecture, anticipate side effects and recognise a deceptively attractive idea will retain all of its value.

The true divide will no longer be simply between those who can code and those who cannot. It will be between those who merely ask AI to produce something and those who know how to guide it, correct it and integrate its work into a coherent technical vision.

This change can be particularly beneficial to open source.

Since my beginnings online in 1994, I have always considered knowledge sharing to be one of the network’s fundamental strengths. The Internet was built by people who published their code, documented their discoveries and enabled others to build on their work.

Artificial intelligence can bring a new dimension to this philosophy. It can make our collective heritage easier to use, help new contributors get started and enable small communities to maintain projects that might otherwise lack resources.

For PrestaShop, it can accelerate the modernisation of a platform that has become very large. For PhenixSuite and thirty bees, it can give new life to architectures derived from PrestaShop 1.6. For module developers, it can make it possible to support more versions, strengthen testing and consider features that were previously too costly to develop.

I nevertheless remain cautious. I have witnessed enough technological changes to know that the initial enthusiasm often hides constraints that only emerge later.

AI raises genuine questions involving confidentiality, intellectual property, security, dependency on a small number of providers and the quality of generated code. It can produce highly convincing errors and give a false sense of mastery to people who do not yet have the knowledge required to identify them.

It must therefore never become an excuse for abandoning good practices. Code reviews, tests, documentation, security work and human validation remain essential.

Despite these reservations, I remain deeply optimistic.

Of all the changes I have witnessed since 1994, artificial intelligence is probably the one with the greatest potential to amplify individual work. It already enables an experienced developer to design, analyse and maintain projects on a scale that would have been difficult to imagine only a few years ago.

I do not see the end of a profession. I see the beginning of a new way of practising it.

Accumulated experience does not become obsolete: it becomes more powerful. The years spent understanding architectures, solving problems and supporting users are precisely what make it possible to use AI with sound judgement.

Having watched the Internet grow from a space reserved for a few enthusiasts into infrastructure that is essential to society, I now feel something similar to its earliest days. So much remains to be invented, the current uses are imperfect and the possibilities appear almost disproportionate.

The difference is that we are no longer starting from zero. We now have more than thirty years of free software, documentation and collective experience behind us.

Artificial intelligence now gives us the means to explore that legacy, understand it more quickly and take it much further.