Building on the Internet Computer: A 2025 Review from the Inside

TL;DR: The SOVERI Review 2025

This is a developer’s candid personal review of building on the Internet Computer throughout 2025, tracing a journey from technical success to deeper reflection and hard-earned lessons:

1. Early Success: We have successfully built and deployed complex applications, including the SIM 3.0 project, proving ICP's technical capabilities and earning IC Alliance membership.
2. The Ownership Crisis: A critical project, SoveriCMS, failed when a managed backend canister (controlled by Caffeine) ran out of compute cycles, and we could not top it up. This exposed a "major contradiction" regarding ownership, control, and accountability within a system built on Web3 principles.
3. Ecosystem and Governance Challenges: I have experienced further disappointment when a technical forum post was deleted and found minimal visible activity or community interest in non-crypto-native, enterprise-focused applications.
4. Conclusion: While the core project remains operational, the accumulation of experiences — especially the lack of clear accountability, the deletion of a technical question, and strategic instability from new announcements — has "gradually diminished enthusiasm and confidence" in the platform's long-term commitment and strategic continuity. The journey is reframed from a marathon to an "ultra-marathon" at Alpe d’Huez.

Read the Extended Version (10 min read)

Writing a yearly review is not something I usually do. This year, however, I felt different - worth capturing in writing, both for myself and for those who may find value in the reflections.

At the end of 2024, we laid the foundation for what became a seven-month project: the deliberate decision to build an application running entirely on the ICP blockchain. That goal set both the technical direction and the learning curve for the year that followed.

Another aspect I want to address in this review relates to my personal perspective as an intrinsically motivated developer and entrepreneur on the Internet Computer. Throughout the year, I became increasingly aware of the contrast between building independently - driven by ownership, responsibility, and long-term commitment - and the ecosystem dynamics shaped by individuals who are directly or indirectly funded by the DFINITY Foundation. This includes influencers, content creators, reviewers and contributors whose involvement is not necessarily grounded in hands-on software development or long-term operational responsibility.

Let me start at the beginning. At the outset, my motivation was simple but strong: to build something meaningful and to stay consistently engaged with the Internet Computer in a practical, hands-on way. As an initial step, I developed a personal fitness-tracking application on the Internet Computer called BeProActive.

The application helped me track bodyweight exercises and align my routine with the WHO recommendation of at least 150 minutes of physical activity per week. While this number may sound modest, I learned that long-term consistency matters far more for health and longevity than sporadic bursts of high-intensity training. By the end of the project, the application was in active use within my family. We used it to track our workouts and share activity through a simple friends feed, which helped maintain motivation by making progress visible and encouraging mutual accountability.

At this point (February 2025), I had accumulated several important learnings - particularly around data migration strategies, efficient data querying with Motoko and the Vue JavaScript framework. Inevitably, new challenges emerged and required practical solutions. Nevertheless, I remained focused and stayed on track, using these experiences to refine both the architecture and my development approach. You can read more about it here.

In May 2025, we had the opportunity to present our work on the SIM 3.0 project built on the Internet Computer to a broader audience in Salzburg, and the response was surprisingly positive. Later that same day, after the event concluded, we traveled to Zurich to attend the 4th anniversary of the Internet Computer. See also my LinkedIn post for more information.

The public presentation of Caffeine and the inspiring atmosphere of the event made the journey worthwhile. It also reinforced our ambition to become part of the new ICP Alliance as a member. We left Zurich significantly more motivated than when we arrived, with renewed confidence in both the technology and the ecosystem. Looking back, there were already subtle signs that a change was coming. At that time, however, I was not able to interpret them correctly. Only months later the full picture became clear.

The first half of the year was dominated by foundational engineering work: data management, indexing strategies, full-text search implementation, navigating canister instruction limits, and identifying the appropriate canister sizes and multi-canister architecture. Progress was incremental, but steady. Piece by piece, each challenge was addressed. By the end, we had achieved all core objectives of the SIM 3.0 project, which is now running in production on the European subnet of the Internet Computer.

I was genuinely proud of the results - especially considering that we were a very small team and that the project ultimately turned out to be far more complex than initially anticipated. With this sense of achievement, I began to promote the project on LinkedIn and through other Internet Computer related social media channels, because it is unique on the Internet Computer. However, the response was minimal. It felt as though there was little interest in a project of this nature. The project also stands for data sovereignty and therefore, the term SOVERI was born.

At that time, we were already operating several websites on the Internet Computer, all deployed on the European subnet, so it was a natural next step to create a dedicated website for the SOVERI project. Until then, these sites were primarily static and increasingly difficult to maintain. This highlighted the need - for both ourselves and our customers - for a content management system (CMS) that would simplify content editing and publishing directly on the Internet Computer.

This led to the creation of SoveriCMS - a concept of a headless CMS with both frontend and backend running on the Internet Computer. Content is managed within the CMS, while the website itself is implemented as a reusable template. During the build process, ASTRO fetches the content from the headless CMS, generates the static site, and deploys it to an asset canister on the Internet Computer, everything done by simply clicking a deploy button. To fully automate this workflow, we integrated GitHub actions to handle the build and deployment process.

As a proud ICP Alliance member, we decided to leverage Caffeine and its vision of the self-writing internet to build this system. From the outset, I was fully aware that this meant working at an early stage of the technology and accepting a certain level of risk. That is part of being an early adopter.

Initially, the goal appeared achievable. With some compromises and 70+ prompts, we reached a functional solution and were genuinely satisfied with the results. However, after continued use, the system began to evolve in an unfavorable direction. Even small changes started to take an unexpectedly long time, and eventually progress stalled altogether. At one point, content data was no longer accessible.

Later, the root cause became clear: the managed backend canister had run out of cycles, a fact I had not recognized at the time, because Caffeine controls apparently everything. This turned into a critical learning moment about ownership and operational control, as well as the practical implications of relying on managed infrastructure in decentralized systems. It highlighted the inherent tension between deeply understanding a system and delegating responsibility to automation - something that works only until it suddenly does not.

A major contradiction became apparent. Web3 is built on the promise of self-ownership. Yet in this case, it appears that someone else was controlling my canister - no different from being dependent on Google Cloud or Azure. This raised a fundamental question: What went wrong?

What happened next was unexpected. I posted the issue on the DFINITY forum, assuming it would be the appropriate place to raise a technical question related to Caffeine. Surprisingly, the response did not come from a technical contributor, but from a forum manager, who informed me that I would not receive any official answer on the forum for a Caffeine related issue.

Fortunately, a community member stepped in and pointed out the underlying issue - the missing cycles - and suggested a possible workaround to recover the data and bring the backend canister back into a functional state. This was genuinely helpful.

Shortly afterward, however, my forum post was deleted by someone. That moment marked a turning point for me. I have been part of the Internet Computer ecosystem for nearly four years, dating back to the Genesis event. During that time, I contributed extensively: publishing free starter kits for frontend development across multiple frameworks, offering and operating a free course

Against that background, seeing a straightforward technical question from an ICP Alliance member removed, while open critique still preserved at the forum, with the suggestion to redirect the discussion to a dedicated Discord channel, was both surprising and disappointing. It prompted deeper reflection on communication channels, governance, and how technical concerns are handled within the ecosystem.

A rather sobering side note: When I posted my question in the official Caffeine Discord channel, a bot immediately responded and appeared to offer help. It quickly became clear, however, that this was a fake account attempting to extract wallet addresses and passwords related to my site.

This experience was a reminder that moments of vulnerability - when you are stuck and asking for help - are precisely when malicious actors try to take advantage. I reported the incident to a channel moderator, who, to my knowledge, removed the bot from the channel shortly afterward.

To keep the SOVERI website online, we ultimately had to rebuild it. We powered ASTRO with a static JSON data file to preserve existing content and protect the work already invested. This approach allowed us to bring the website back online in its current form - the version you are reading now.

The SoveriCMS implementation based on Caffeine has since been closed. Over the coming months, we will rebuild the system using a more traditional, human-driven approach - augmented by AI where it adds value, but never without clear ownership and control. This ensures that learning remains part of the process and that responsibility is always explicit.

This experience highlights a broader lesson: When control is delegated to automated systems without clear accountability, even small issues can escalate quickly. This is not a critique of Caffeine specifically, but a general warning relevant to any system where responsibility becomes opaque. If no one is accountable, no one can act when things go wrong. That is a risk worth reflecting on carefully.

By autumn, I was approached with an invitation to join a group of Internet Computer reviewers. This coincided with the Voting Neuron Grants (VNG) 2 program of the DFINITY foundation, which still had roughly two months remaining.

The idea was to review ICP proposals collectively in exchange for compensation. The person who contacted me explained to me later on that they were looking to fill a gap in the group and saw this as a mutually beneficial opportunity.

This business was entirely new to me, and after some careful consideration, I saw it as an opportunity to deepen my understanding of the Internet Computer - less for the financial aspect and more for the learning process, the discussions, and the people involved.

However, once I became more involved, it became clear that the group was primarily economically driven, especially after a group meeting. There is nothing inherently wrong with that, but it was not aligned with what I was personally looking for. For that reason, and a few others, I ultimately decided to step away from the group.

Although I attempted to introduce a new perspective - by visualizing the review process to improve transparency and by laying a foundation for future reviewers - I also implemented a simple, time-saving notification service for newly submitted proposals. This eliminated the need to constantly monitor the ICP dashboard during the three-day review window.

At the same time, I noticed a clear contradiction: despite the significant amount of money involved, the review process itself lacked structure and professionalism.

My impression was also that the primary motivation was to fill a gap quickly so the group could participate financially. Beyond the person who initially invited me, very little information about the actual review process was shared. In the end, there was limited interest within the group in adopting or building upon the improvements I had proposed.

Fast forward to today: Proposal Review Season 3 has been announced (16.12.25) by the DFINITY Foundation. This time, however, only individual reviews are accepted - group reviews are no longer permitted - and the scope of topics has been significantly reduced.

Another new experience for me was the realization that decisions could affect one’s work and future in certain areas are made through voting by people I do not know. As someone who is neither a politician nor accustomed to such governance dynamics, this required adjustment. Ultimately, I chose to leave the group on my own terms, carrying with me mixed but valuable reflections.

Around this time, SIM 3.0 was facing its first real production test cases. We were able to adopt one or two components successfully, and technically everything remained on track for a production start in January 2026. However, the positive momentum began to fade.

There was little to no visible activity or communication surrounding the newly announced ICP Alliance. I reached out to several contacts, but the response was minimal, and interest seemed noticeably lower than expected. A proposed ICP Alliance Summit scheduled for autumn was ultimately not held, and no further updates were shared - despite the introduction in a mandatory workshop that participants were required to complete in order to qualify for the summit.

In early November 2025, a Medium article by Dominic Williams, the chief architect and visionary behind the Internet Computer, was published. It introduced concepts such as Internet Computer 2.0, DFINITY 2.0, and Caffeine as a paradigm for self-writing applications. This announcement left many speechless, myself included. The article highlighted how uncertain and fast-moving the future of the platform can be. It marked a transition from a research-oriented organization to a new, more commercially driven spin-off.

While the piece aimed to communicate a positive and ambitious outlook, it did not resonate in the same way for me. Instead, it reinforced a growing sense of instability - particularly for those building long-term, production-grade systems and making strategic commitments based on architectural continuity.

The community's scant attention to non-crypto-native developments represents a significant blind spot, especially concerning enterprise solutions and business applications. The kind of work that real companies actually need, the "last mile" towards end users, often seemed to be of limited interest to the broader ecosystem.

During this period, it also became evident that many people had been laid off from DFINITY. Some were able to communicate their departure publicly through LinkedIn posts or statements on X, while others were not. In several cases, I only became aware that individuals were no longer part of the organization when emails bounced back with messages from the postmaster indicating that the address was no longer valid.

This led me to a fundamental question: Is anyone asking how intrinsically motivated contributors within the community actually feel? Communication matters, and the accumulation of these experiences gradually diminished my enthusiasm and confidence in the Internet Computer.

Of course, there is no formal obligation to communicate with the community more than it is currently done. However, trust is a foundational principle of blockchain technology - and once that trust begins to erode, the question arises: What remains?

People close to me began asking difficult but valid questions: Should we continue building on the Internet Computer at all? Is it realistic to commit to this platform over the long term?

I recall a statement from a former DFINITY CTO made during a public panel discussion: "This is not a sprint; it is a marathon." I would slightly reframe that analogy. From my perspective, this feels less like a marathon and more like a bike-ultra-marathon - with 10,000 meters of elevation gain - and we are only around kilometer 30 and 500 meters of elevation gain. It feels like a Grand Tour stage in cycling - climbing Alpe d’Huez twice.

As I write this 2025 review, our project is fully operational on the European subnet of the Internet Computer - the subnet with the highest instruction throughput - and it continues to run reliably in production. This is an important and positive outcome that deserves explicit recognition. Certainly, update calls may take an extra second when required - but security always comes at a cost.

However, this high instruction throughput is not grounded in the fact that many European canister owners deploy on this subnet. The elevated throughput exists largely because of the sustained load generated by a single major project: BOB.

So how do I conclude such a long and personal review? I believe it is essential to speak openly about one’s experiences and perspectives. This text is published on the Internet Computer, and it cannot simply be removed by a community or forum manager. Whether this makes some people uncomfortable or resonates with others is not my primary concern. These are my reflections.

I am not attempting to judge individuals or institutions. I am describing my own experiences and feelings, openly and honestly, in a way that I believe is both fair and appropriate for public discourse.

This is also a conversation with myself and with people like me - those who invest significant time and personal resources into learning, understanding, and trying to move things forward. Often, this effort happens quietly, with little visibility, recognition, or compensation.

I believe this is a necessary conversation

As this year comes to an end, I still stay confident in my own path. I focus on building, learning, and taking responsibility for what I create - openly and without hiding behind acronyms or anonymous email addresses. I believe that consistent hard work, real understanding, and personal accountability will always prevail in the long run.

Wishing you a happy, healthy, and inspiring New Year.