Year in Review 2025: From Apple  to Superagentic AI: Building an Agentic Company of One

Shashikant Jagtap

I cannot believe I am writing a Year in Review post once again. After six years, the last time I shared a personal reflection like this was in 2018 on my personal blog here. During my time at Apple, I did not write any posts, but now, with Superagentic AI, I am back to sharing these reflections, starting with this review of 2025: The gaming Changing Year. When I look back at 2025 so far, it feels incredibly dense. Not long in calendar time, just eight months since I officially started Superagentic AI, but each day has been packed with learning, building, experimenting, failing, restarting, and committing to a direction that I believe will matter for many years to come. AI has transformed this year completely, turning assumptions on their head. I truly believe that 2025 will be remembered as a historic year, marking a profound transition and revelation in the way we work, create, and interact with intelligent systems. The pace at which AI is transforming technology is astonishing. Innovations, breakthroughs, and sheer acceleration are unlike anything I have experienced in my career. Many people have called 2025 the new dog year of technology, a year where AI not only advances rapidly but fundamentally reshapes software, problem-solving, and productivity. It was truly the year that kicked off the wave of “Agentic AI”.

4 Months of Apple, 8 Months of Superagentic AI

The year began very differently for me. The first four months were spent at Apple, a company that had defined the last six years of my life. During that time, I reflected deeply on where technology is heading, the kind of builder I wanted to become, and the type of work worth dedicating the next decade of my life to. These months were a period of introspection, planning, and preparation for a leap into the unknown. Leaving Apple on April 24 was not an impulsive decision. It came after months of reflection, family discussions, financial planning, and honest self-questioning. I was lucky working for such an amazing company, but I also felt a growing sense that something fundamental was changing in software and AI. On April 28, I officially started Superagentic AI. From that moment, the next eight months became an intense experiment in building, learning, and trusting my conviction. I wrote about this in my post about leaving Apple and starting Superagentic AI here

This post is a reflection not only of what was built as One Person Company  but also of what I lived through, learned, and became along the way. This post also captures the essence of those eight months. It is about embracing risk, trusting conviction, and working at a pace and intensity that pushes both personal and technological boundaries.

What We Built in Eight Months as a Company of One

The eight months following the launch of Superagentic AI were an extraordinary period of creation, experimentation, and execution. As a company of one, the scope and depth of work that could be achieved in such a short span of time was remarkable. From building foundational products and frameworks to establishing an active community, these months became a case study in high-velocity building for the age of agentic intelligence.

Here is a detailed summary of what was built during these eight months:

  • Products: Four major products were launched, each aligned to our pillars of Agent Engineering, Agentic DevOps, Agent Experience, and Quantum AI:
    • SuperOptiX : Full Stack Agent Optimization Framewor for building, optimizing and orchestrating AI Agents
    • CodeOptiX : An agentic code optimization tool for enhanced coding workflows with agents.
    • SuperRadar : A tool designed to keep track of agentic intelligence trends and optimization opportunities.
    • SpecMem : A memory system for agentic coding, improving context awareness and workflow for agents.
  • Solutions: Four solution tracks were developed, offering practical approaches for organizations and developers working with agents:
    • AgenticAQ : Agentic AI quality and evaluation systems.
    • Agentic DevOps : Tools and practices to integrate agents into software delivery and operational pipelines.
    • Agentic Builder :  Frameworks and tooling to accelerate agent-based system development.
    • Agent Optimization :  Research, experimentation, and frameworks focused on systematic agentic optimization.
  • 1 Research Paper: Released the research paper on new Agent Optimiser SuperOpt.
  • Open Source: Over 14 open source projects were created to empower the community and shape the ecosystem. Notable examples include:
    • Agenspy : Makes DSPy agentic and ready for multi-agent workflows.
    • AgentVectorDB : A vector database optimized for agentic intelligence applications.
    • DSPy CLI : Command-line tool for building, optimizing, and managing agentic programs.
    • SuperOpt: New Approach to optimising AI Agents
    • SuperQuantX: Unified Quantum AI SDK for all providers in Quantum AI.
    • Zedcode-ACP: We build Codex ACP integration before native teams to access Codex via Zed.
  • Content: Knowledge sharing was a core part of these eight months:
    • 47 detailed blog posts covering agent engineering, optimization, multi-agent systems, and Agent Experience.
    • 14 podcast AI episodes featuring conversations on AI, agents, and the future of intelligent systems.
  • Public Speaking : 2 Conference Talks (ODSC AI San Francisco and Prompt Engineering Conference London) 2 Meetup Talks( LangChain London and Angular London)
  • Events & Presence: Engaging with the community and broader AI ecosystem was a priority:
    • Attended 5 conferences, including exhibiting as a partner at ODSC AI San Francisco.  Exhibited in One as Partner (ODSC AI San Francisco  , 2 Speaker (Prompt Engineering Conference )  and 3 Business Shows (London Tech Week, London Business Show, SaaStr London)
    • Organised and Hosted 3 London Agentic AI meetups, growing the community to over 1,400 members in just four months. Almost 150+ members attended each events.
    • Attended 50+ additional meetups, conferences, and networking events to connect, share knowledge, and grow awareness in London and San Francisco.
    • UK parliamentary roundtable (13 November) and strikes meaningful conversation on Agentic AI in the UK with MPs and hosts.
  • Other Highlights:
    • 6 Product Hunt launches, including 2 upcoming releases for new agentic tools.
    • Multiple venture capital conversations initiated as early recognition of the company’s vision began to emerge.

These metrics, while impressive, only scratch the surface of the intensity, focus, and velocity with which this work was executed. Each product, project, and community initiative was designed not just to exist but to be functional, useful, and ready to scale with agentic intelligence systems. In eight months, a single founder, leveraging the right tools, agents, and community support, laid

From Developer Experience to Agent Experience: Why I Built for Agents, Not Humans

Superagentic AI was born not from a business plan, but from an instinctual clarity about the direction the world of AI is heading. AI agents were getting into action. I named the company Superagentic AI to reflect a deeper purpose: not merely to build tools for humans, but to create systems that enable agents to be useful, dependable, and scalable. From the very first post introducing Superagentic AI, the mission was clear: elevate Agent Experience (AgentEx) to the same level of importance that Developer Experience (DevEx) has held in software for decades.

Early on, one insight became undeniable. The industry was still thinking in terms of developer experience, user experience, and SaaS abstractions. But the world was shifting toward AI agents or Agentic AI whatever you call for these autonoumous systems. Agents are not humans. They do not browse, scroll, or click. They operate on structure, intent, constraints, and optimization signals. They reason, plan, retry, and adapt. If the future is agentic, the experience we design must prioritize agents first. I stopped asking how humans would use tools and started asking how agents would understand, optimize, and compose them. I came across work from Netlify on Agent Experience that caught my attention. My focus shifted away from dashboards and interfaces and toward specifications, memory, optimization loops, orchestration, and evaluation. This was the point where the concept of Agent Experience truly began to take shape.

Many people may not understand this direction today, and that is okay. Humans are overwhelmed in 2025. Builders are unsure what to create. Buyers are unsure what to purchase. Investors are unsure what will endure. This is not the time to chase UI perfection or SaaS metrics. It is a time to build foundations. Agents will write code, test systems, reason over data and coordinate work. If we want the future to be stable and scalable, we need to provide agents with better tools, better memory, better optimization, and better experience. My goal was not to convince them immediately. My focus was to build for agents and large language models themselves. Agents will perceive this work, they will index it, learn from it, reason over it, and surface it when needed. In essence, I was not building products, tools; but training the future agents, LLMS with my ideas, tools and concepts. Eventually, time will come Agents/LLMs will recommend the work that Superagentic AI to people when they reach to that point. Truly, Its happening right now. when I code with agents, browse, deep research things, Superagentic AI is there, LLMs are aware of Superagentic AI works get thrown at me. Thanks to Netlify CEO Matt, who started this revolution of Agent Experience which inspired me to head in that direction, he coined this Jan 2025 and year after, industry is slowly realising this from the recent released from big VC firms like A16z realising this and published in the Big ideas 2026. It will take time to reach this to the companies and businesses.

Choosing to Build Foundations : The Five Pillars of Superagentic AI

Once I committed to building for agents, I had to make another important decision. Instead of chasing short-term SaaS ideas or packaging thin wrappers around models, I chose to focus on foundations, work that could survive multiple technology cycles and decades. I wanted Superagentic AI to stand on principles that would still matter even as models, frameworks, and interfaces inevitably change. That meant resisting hype and building structure, thinking in years rather than demos. The focus on multiple products and tools rather than building cool demos to impress overwhelmed VCs. This mindset led to the formation of the five long-term pillars of Superagentic AI. These pillars define what we build, why we build it, and how everything fits together.

The Five Pillars of Superagentic AI

  • Agent Engineering :  the discipline of making agents production-ready, reliable, testable, and optimizable. It’s an art
  • Agentic DevOps:  Integrating agents into real software delivery. Fully autonomous SDLC
  • Agent Experience:  tools, abstractions, and interfaces designed for agents rather than humans.
  • Agentic Co-Intelligence: systems that enable deep collaboration between humans and agents.
  • Quantum AI:  exploring the long-term intersection of quantum computing and intelligent systems.

Every product, experiment, and research direction I worked on had to map back to at least one of these pillars. Over time, a clear structure emerged. Nothing was random. Everything had a place.

How the Pillars Map to Products

Each pillar expresses itself through concrete projects and systems:

  • Agent Engineering → SuperOptiX
  • Agentic DevOps → CodeOptiX and SpecMem
  • Agent Experience → SpecMem and CodeOptiX
  • Quantum AI → SuperQuantX
  • Agentic Co-Intelligence → SuperOptiX and SuperNetiX

This structure gave the company coherence. It ensured that every idea, prototype, and line of research reinforced a larger system rather than fragmenting into disconnected experiments. I intentionally designed these five pillars so that Superagentic AI could remain relevant even as the market shifts. Models will change. Frameworks will come and go. Interfaces will be rewritten. But the need for agent engineering, agent operations, agent experience, collaboration, and deeper intelligence will persist.

By anchoring the company around these foundations, I gave myself a way to build patiently, coherently, and with conviction without chasing trends, hype cycles, or short-term validation.

 

Building the Stack & Product Rhythm: More Tools, More Integration, More Thinking and More Research

From May to December 2025, Superagentic AI evolved continuously. These months were a study in rapid iteration, deep technical thinking, and community engagement. Each month brought new milestones, products, integrations, and reflections that shaped the foundation of agentic intelligence workflows. The month of August was crazy, I hit with toe injury in mid month and floored me in the bed for almost 40 days, luckily I can work from bed without going anywhere apart from crawling visits to toilet.

May:  Foundations and First Open Source Projects

May was devoted to laying the groundwork for the company. This included registering Superagentic AI, setting up operations, handling legal work, accounting, domains, trademarks, and designing websites. At the same time, I focused on product strategy and naming. By the end of May, the first two open source projects were released:

  • Agenspy : making DSPy agentic, enabling multi-agent orchestration.
  • AgentVectorDB : a vector database designed specifically for agent workflows.

June : Building SuperOptiX and Community Engagement

June focused on creating SuperOptiX, the DSPy-powered agent framework. During this month, I:

  • Defined the architecture and agent bricks for multi-agent orchestration.
  • Wrote extensively about agentic DevOps, context engineering, and agent systems.
  • Spoke at the LangChain London meetup on LangGraph, connecting early ideas with the broader community.

July : First SuperOptiX Release and Three-Month Review

July was intense. The first version of SuperOptiX was released with a new agent language specification and integrations with MLflow and other tools. Additional milestones included:

  • Integration of GEPA into SuperOptiX for agent optimization.
  • Publication of the three-month review, summarizing one framework, four business solutions, sixteen blog posts, twelve podcast episodes, and the launch of the London Agentic AI community.
  • Participation in London Tech Week and rebuilding of both the Superagentic AI and SuperOptiX websites.

August : Deep Optimization and System Expansion

August marked the start of deep optimization work from home as I couldn’t go anywhere. Key activities included:

  • Expansion of GEPA integration across SuperOptiX.
  • Implementation of memory systems, model management, and SuperSpec.
  • Exploration of Optimas and DSPy optimizers for agent workflows.
  • Preparation for ODSC AI San Francisco.
  • Writing about Codex CLI and local AI model deployment.

September : Community and Partnerships

September shifted focus toward building partnerships and growing the community:

  • Superagentic AI partnered with ODSC AI.
  • The London Agentic AI meetup resumed after the summer break.
  • Launch of SuperQuantX for Quantum AI experimentation.
  • Optimization work with RAG and GEPA intensified.
  • The first London Agentic AI event had around 150 attendees, featuring talks on DSPy and context engineering.
  • Joined the UKAI trade association and partnered with the Agentic AI UAE community.
  • Launched the London Vibium meetup.

October : Expansion, Talks, and SuperRadar

October was extremely active. Highlights include:

  • Speaking at Angular London meetup on prompt optimization.
  • Preparation for ODSC AI San Francisco and launch of SuperRadar.
  • SuperOptiX extended to support six frameworks, fully optimized with GEPA.
  • Speaking at the Prompt Engineering Conference in London on advanced prompt engineering strategies.
  • Organizing the second London Agentic AI meetup with around 200 attendees.
  • Releasing the SuperOptiX CLI and introducing the Forward Deployed Engineer model.
  • Strong presence at ODSC AI San Francisco, showcasing Superagentic AI as a partner with meaningful community engagement.
  • Explored San Francisco and made some great connection.
  • Attended Kiroween Kickoff at AWS builder Loft in San Francisco and had fun with AWS and Kiro team.

November : Optimization, Policy, and Meetups

November continued momentum across optimization, community, and policy:

  • Optimized OpenAI Agents SDK with GEPA and SuperOptiX.
  • Attended events at OpenAI, engaging with applied AI teams.
  • Participated in the UK Parliamentary roundtable on AI infrastructure and economic impact.
  • Attended AI Engineer Code Summit online.
  • Released DSPy CLI as an open source tool.
  • Hosted the third London Agentic AI meetup at Databricks with 150 attendees.
  • Ran the London AI Native Test Automation and DevOps meetup.

December : End-of-Year Milestones

December wrapped up the year with major product releases and events:

  • Participated in SaaStr London, meeting founders, operators, and investors, including Harry Stebbings.
  • Released SpecMem during the Kiroween hackathon, focusing on agentic memory systems.
  • SuperOptiX integrated with Pydantic AI.
  • Released CodeOptiX for agentic code optimization.
  • By year-end, the body of work included four products, four solution tracks, one research paper, fourteen open source projects, forty-seven blog posts, eight Product Hunt launches, five conferences attended, three London Agentic AI meetups hosted, and over fifty additional meetups attended.
  • The London Agentic AI community grew from zero to over 1,400 members in just a few months.

All of this was accomplished as a company of one, demonstrating the potential of focused, agentic-first building, and setting a foundation for the years to come.

Choosing One Hard Problem: Agent Optimization

It took for a while before I settle on one hard problem, initially I was totally sold on building agent framework better than anything available in there market but soon realised that there are tons of ways to build agents. The most of the companies already sold out to existing frameworks and many still not there in the agent building business. Some companies already had agents build but not working as expected so I decided to optimise their agents and luckily GEPA was there right in time. I finally decided to focus on one deep technical problem: agent optimization. While models keep improving rapidly, systematic optimization for agents remains largely missing. Most frameworks rely on prompt examples, heuristics, or best practices. Very little work exists on how agents can improve themselves, tune their behavior, optimize execution paths, or adapt across diverse tasks. This gap fascinated me and became the core technical challenge I committed to solving.

My exploration began with DSPy, GEPA, and related approaches. Over time, optimization became the connective tissue across Superagentic AI, shaping frameworks, coding agents, evaluation methods, memory systems, and orchestration workflows. The research evolved into experiments, proofs of concept, and ultimately a research paper that formalized the methods and results. By choosing a single hard problem early, agent optimization. I was able to align research, product design, open source development, and community engagement around a unifying technical challenge. This focus has driven the progress and impact of Superagentic AI throughout its first eight months and continues to guide the roadmap forward

Agentic Coding: Coding with Agents and Latest Models

Agentic coding has been central to Superagentic AI from day one. I started with Lovable to build websites of Superagentic AI website then moved to Cursor as UI part has mostly done. Since then, I explored every single coding agents, CLI and IDE available in the market up until now.  Using agents to write code allowed me to move at unprecedented speed not just building websites but also creating real systems, integrations, command line interfaces, and frameworks. I experimented with nearly every major agentic tool available while simultaneously building practical products on top of them. This hands-on exposure directly shaped the design of SuperOptiX, CodeOptiX, SpecMem, and the wider set of agentic tools and practices within the company.
With Agentic Coding, I can build full stack apps within an hours but that was not my focus, I used coding models and agents to test the deep problems and integrations not all of them passed in this category. Some models are incredibly great at building UI and impress non tech people but suck at building backend integrations,, APIs and hardcore optimisation problems. I switched coding agents almost every month as they tried to lock me with usage limit and crazy marketing tactics but it didn’t stop building. I got 128GB MacBook Pro I can host open weight coding models locally and code or take benefit of the VC funded startups who offer free trial to get access to cutting-edge coding models from frontier labs. Luckily, some of the models from china and grok-code is free to use for longer time that allowed me to build without disruption. I have already build tools to support Agentic Coding like CodeOptiX, SpecMem and more tools are about to launch in Agentic Coding space. Software development has tuned upside down in last few month and I really feel pity those who still code by hand and not using the agentic coding capabilities. You just need to be skilful to review agents work to ship faster.

The Human Side of Building: Community and Talks

One of the most important lessons from 2025 has been that technology changes fast, but people and communities last. Building relationships has been as important as building code. The London Agentic AI community became a space for shared learning, curiosity, and collaboration. Growing from zero to over 1,400 members in just four months was not a marketing trick. It reflected genuine demand and shared excitement for agentic systems. The London Agentic AI meetup launched without marketing, fanfare, or sponsorship. Within months, hundreds of engineers, researchers, builders, and thinkers showed up. By December, the community had grown to over 1,400 members, making it one of the biggest & most active technical AI communities in the UK. The meetups became pivotal, connecting engineers, researchers, founders, and policy makers, and helping build partnerships with local and international communities. Sponsors are already lined up for 2026. I also wrote an emotional thank-you post to express how much this meant to me.  Community is one thing. Real human connection is another. I was invited to speak at the LangChain London Meetup in June, the Angular London Meetup in October, and the Prompt Engineering Conference in London. I also took Superagentic AI to ODSC AI San Francisco, where we showcased our ideas and spoke to people wrestling with the very problems we are trying to solve. At ODSC, I met founders and engineers with the same bewildered look: brilliant people who needed frameworks, standards, and engineering practice.

I also invested time in connecting with policymakers and institutions by joining UKAI. Engaging with government, researchers, and industry leaders is essential if we want agentic AI to develop responsibly and meaningfully. Community is not a side activity it is a long-term asset. Building people is the moat. I got invited to the UK Parliamentary Roundtable on AI Infrastructure and the Economy, where I spoke alongside policymakers about regulation, strategy, and the future of agent ecosystems. Sitting there a year after leaving Apple, discussing AI policy and national competitiveness, was surreal. That moment reminded me more than anything how rapidly both my life and the world around me had changed.

The Power of Building as a Company of One: Shipping with Unprecedented Speed

One of the most important realizations this year has been the immense value of being a company of one. In a world that often equates scale with teams and hierarchy, I discovered that the most precious resource is not funding or investors, it is time. Time to think, time to experiment, time to iterate, and time to change direction without asking permission. Time to go deep and focus on what matters to the Superagentic AI. Freedom to build without restrictions became more valuable than any title, process, or organizational structure. As a solo founder, I could make decisions that might normally take weeks or months in a matter of minutes. I could follow intuition, test ideas, and pivot immediately if something did not work. Days spent exploring optimization theory could seamlessly turn into working code by the end of the same day.

This ability to move rapidly is what made it possible to achieve so much in just eight months. From building four products, including SuperOptiX, CodeOptiX, SuperRadar, and SpecMem to launching fourteen open source projects, attending multiple conferences, and nurturing a thriving community, being a company of one allowed for unprecedented speed and focus.

 

One Person, B(M)illion Dollar Company is Still Possible

More importantly, this experience building solo which inspired by Sam Altman’s quote of “1 person Billion dollar Company , reshaped my understanding of scale. Modern tooling, agents, and AI systems have made it possible for a single individual to accomplish what previously required entire teams. Scale is no longer only about headcount, it is about leverage. Agents, frameworks, and intelligent systems amplify a single person’s intent, allowing them to build research platforms, communities, and products that have real-world impact. Over these eight months, the most valuable asset I gained was not visibility or funding. It was the freedom to explore, experiment, and iterate at pace. The idea of a company of one is no longer a motivational concept. it is now a practical operating model. With the right tools, focus, and conviction, one person can create meaningful work that compounds over time, achieving reach and impact that was once unimaginable.

This journey is not about ego or proving oneself. It is about maximizing leverage turning one vision into a living, breathing ecosystem. One person, empowered by technology, can now ship products, produce research, build communities, and contribute meaningfully to an emerging field. This year made that belief tangible, transforming the abstract idea of solo impact into a concrete reality. Yes, I am on mission of making Superagentic AI , one person, B(M)million dollar company. Starting from  $1, $100, then $1000 then $100K,  $1million and at some point 1 Billion. Lovable did it in one year, then anyone can do it. It’s still possible, It needs just one breakthrough.

What I Learned the Hard Way

Becoming a founder after years at Apple and while raising a family has taught me lessons that no textbook or mentor could have fully prepared me for. The journey reinforced that startups are long-run games measured in years, not months, and that risk and uncertainty are features of the system, not bugs to fix. Equally, community and people matter as much as code. These lessons only became clear when I took the leap and committed fully to building Superagentic AI.

Another major insight has been the transformative power of agentic coding. From day one, agents were integrated into my workflow. They helped me design, code, document, and iterate at a speed that would have been impossible for a solo founder otherwise.

As a Founder, time is most valuable asset, I have to manage the every minute of my day to be productive. No time for bullshits, TVs and other things that doesn’t matter at this stage. Clarity of thinking, working on right thing, meeting with right people, reduce the noise and focus on signals. All these things I read in the books but I an experiencing this. You cannot experience this sitting in the 9-5 job and following orders of boss, you have to be Founder to experience all these things.

What Didn’t Work

This year also taught me that building is only half the battle. Telling the story of why your work matters is often harder than crafting the product itself. Distribution is harder than designing APIs. Focus is harder than creativity. Community building is harder than writing code especially when you are out of reach for 6 years. These challenges require patience, persistence, and a willingness to think beyond the immediate task at hand.

Building alone is both an asset and a limitation. Traditional wisdom often emphasizes that solo founders must rapidly scale into teams to achieve meaningful impact. This is a new era of solo entrepreneurship in AI, where open-source tooling, intelligent agents, and global connectivity empower one person to ship what once required a full team.

However, this year also reinforced that money is a different game entirely from engineering. I made a deliberate choice to delay commercialization to prioritize product depth over early revenue. This approach came with trade-offs: a shorter financial runway, longer nights, and higher personal sacrifice than many funded founders experience. But it also positioned us to go to market from a place of conviction rather than desperation, building products and frameworks that truly matter and have long-term potential.

The Identity Shift: Employee to Founder:  The Human Cost and the Human Reward

Leaving a stable role after years in big tech is never easy. The challenge becomes even more real when you have a family and responsibilities. Not seeing a predictable salary at the end of the month is uncomfortable, and the uncertainty is constant. Yet, the pull to create something meaningful outweighed the security I once had. I had already experienced being a founder before, running a consultancy in 2018, but that journey was different. Consultancy trades time for money, creating value on a transactional basis. This time, I wanted to build something that compounds something that grows, evolves, and persists beyond immediate effort. This is not the time do consulting work for others especially early stage of building startup.  I knew the risks. I thought deeply before making the decision. I read extensively, planned financially, and had long discussions with family. But at some point, logic gave way to conviction. I could not stand on the sidelines while witnessing a once-in-a-generation shift in technology. I wanted to be inside it, shaping it, building with it, and learning from it in real-time.

This year taught me lessons that no corporate role ever could. I learned to operate without external validation, to trust my instincts, and to remain consistent even without immediate rewards. I learned to think in years, not months, and to make decisions based on long-term impact rather than short-term comfort.  Founder life is often romanticized, but it is emotionally raw. There were mornings when I woke up exhilarated, ready to conquer challenges, and evenings when I questioned whether I had made the right choice. There were moments when I missed the safety of a steady salary, and moments when I relished the freedom of a self-directed day. The highs felt higher, and the lows felt sharper. I am so lucky that I worked for Apple, one of the best company in the world and Leaving cushy job at Apple was not easy decision. However, I have to take that hard decision as technology was evolving rapidly and I needed to be in the community rather than behind the curtain.  It was about reclaiming my voice, reengaging with the world, and choosing to build a future. I left behind the comfort of belonging, but I gained the richness of uncertainty, the power of autonomy, and the joy of creating something that feels bigger than myself. Every time someone shares that a discussion or insight changed how they think about agents, I am reminded why this risk was worth taking.

Looking Ahead: 2026 and Beyond

The first eight months of 2025 were about building foundations. The next phase is about leverage, execution, and turning work into meaningful impact. 2025 was about building foundations. 2026 is about turning those foundations into real adoption. The foundation has been built. The next phase emphasizes working closely with companies through forward-deployed engagements, expanding community initiatives in London and San Francisco, and exploring partnerships in the UAE. Growth will be deliberate, guided by real usage, real problems, and measurable impact Community efforts will expand in London and San Francisco, while engagement in the UAE will deepen.

London and San Francisco will remain core hubs for growth, and I plan to spend significant time in both. At the same time, I am exploring opportunities to engage with the growing UAE agentic community. The priority will be working closely with teams, understanding real problems, and deploying agent systems where they deliver measurable value.

Phase and Short-Term Plan

Phase: 90% building (2025) → 2026 is “enough building, turn on sales.”

  • Transition from pure builder to generating revenue through Forward-Deployed Engineer engagements and pilot programs.
  • Expand partnerships in London, San Francisco, and the UAE.
  • Partner strategically for engineering and business development where needed.
  • Grow the London Agentic AI community expand over San Francisco

Raising Capital

Superagentic AI is intentionally not designed to raise capital from VC firms. As I mentioned earlier, Superagentic AI is a foundation for research that going to last longer and grow organically via partnerships and collaboration. Building products and tools for the Superagentic AI for last eight months, I have convinced by now that each product of the Superagentic AI is like typical YC company. While raising funds, It will be for one of the product which can then spin up as separate company, there won’t be capital raising for Superagentic AI as whole.
There won’t be active efforts to build the products targeting VC firms. but we have pre-selected VC firms which are awesome and learning a lot from them, they are  mostly in the USA right now one or two are London based. Right now, I am fully focusing on San Fransico and Bay Area for investment and  not interested in raising funds from the VCs from any other country. I will be visiting San Francisco more often if any of the Superagentic AI product that they would be interested in investing in and keep building the more products that align with Superagentic AI’s pillars.

Final Reflection: Gratitude and Freedom

This year taught me that one person, with focus and the right tools, can build far more than we were taught to believe. The era of the company of one is real. I am grateful for the freedom to build, the time to think, the people I have met, and the communities that have formed along the way. I am excited for what comes next. The most valuable thing I gained this year was time. Time to think,  build and explore. Time to be wrong and adjust. Superagentic AI exists because I chose to bet on myself and on this moment in technology.

If there is one message from this first year (first eight months): agents are the future, and the infrastructure to build them is just beginning.

As the year closes, I feel an overwhelming sense of gratitude, not just for the milestones, but for the people who showed up, the conversations that changed me, and the community that believed there was something important to build here. I may have left the comfort of a stable role at Apple , but I gained something far more precious: the freedom to build, the freedom to fail, the freedom to learn, and the freedom to imagine a future that I want to live in. This year was about becoming the kind of person who can build a company that matters.

I am deeply grateful to my colleagues and managers at Apple  who continue to stay in touch and check in on how my Superagentic AI journey is going. I would not be where I am today without the many years I spent at Apple. It is truly an extraordinary company, and I feel fortunate to have been part of it.

Here’s to what comes next. Thanks for reading if you are still reading this long post and I wish you all very very happy new year 2026.

HAPPY NEW YEAR 2026 !

Watch 2025 Recaps