By Billy P. · Founder Notes #4.
Over the last few days I have been doing something that is probably not very exciting from the outside.
Breaking things.
More specifically, I have been working through the Foundation Stability Audit for ISAC. Stage 35 of the roadmap is already complete, but I made the decision not to move straight into Stage 36. Before I start adding external providers, cloud models and the next layer of ISAC, I want to know that the foundation underneath it is actually solid. Not “it seems fine.” Not “the tests passed once.” Actually solid.
The sequence right now looks like this. Finish Phase 2. Then the Phase 2 source-control closeout. Then Phase 3. Then the Phase 3 source-control closeout. Then the full post-fix Foundation revalidation. Then the Foundation Stable certification. Then the benchmark package. Then the FOUNDATION STABLE LIVE DEMO. And only after all of that do I move into Stage 36.
That live demo is becoming quite an important milestone for me.
Why this demo matters
Because up until now, a lot of what I have been building has existed in code, architecture documents, tests, development logs, and the occasional screenshot. The Foundation Stable demo is where I want to put the system in front of people and actually show it working.
No slide saying what it should do. No architecture diagram saying what it could do.
The software running.
GCF 2.0 is already done
There is also an important distinction here that I probably have not explained clearly enough.
GCF 2.0 is already done. GCF is the Governed Cognitive Framework underneath ISAC. It provides the cognitive framework around things such as planning, governance, prediction, world modelling, cognitive economy, recursion, background cognition, and governed learning. ISAC is the AI system being built on top of it.
So when I talk about Stage 35, Stage 36, or the Foundation Stability Audit, I am talking about the ISAC implementation roadmap, not still trying to invent GCF.
That matters because GCF is now the baseline. The point of the current work is to prove that ISAC can use that framework reliably before I start stacking much more on top of it.
Stage 36 is going to be a big jump
Stage 36 is where external and custom model providers start entering the picture.
OpenAI. Anthropic. Custom endpoints. Self-hosted providers. BYOK. Cost controls. Privacy policies. Fallback between providers.
That is a significant increase in complexity. I do not want to introduce that complexity onto a foundation I have not finished trying to break first.
So why am I raising £250k
The short version is the work is about to get bigger than what founder-funding can carry on its own.
GCF FramWorks LTD is pre-launch and pre-revenue. The architecture is currently at Stage 35 of a defined 55-stage roadmap. The current checkpoint is closing Stage 35 with benchmark evidence, resource safety, regression tests, documentation, and merge approval. After that comes Stage 36, then 42, then 49, then 53, then the release of validated ISAC Core v1.0 at Stage 54-55.
Some of that work can be done solo. Some of it cannot. Doing it properly is going to require external review, more compute than I can rent by the hour, real productisation work, and at least one structured pilot with an outside user. None of that fits inside a personal budget.
The £250,000 is the difference between “founder-funded technical prototype” and “validated AI architecture with pilot-ready integration, benchmark evidence, and security review.” That is the honest framing.
The current proposal is £250,000 for 15% equity. That implies a post-money valuation of roughly £1.67 million and therefore a pre-money valuation of roughly £1.42 million. SEIS eligibility is subject to confirmation.
I want to be clear about what that number is. It is not me claiming that somebody has independently valued the company at exactly £1.42 million. It is not £1.42 million sitting on a balance sheet. And it is not based on current revenue, because the company is currently pre-revenue. It is the implied valuation created by the proposed investment terms.
The more important question is therefore not where did £1.42 million come from. It is what does the £250,000 actually get the company to. That is the part I think founders sometimes explain badly.
What does the money actually buy
The pitch deck puts it in one line that I think is the right framing. Capital is tied to measurable technical and commercial milestones rather than feature count. The valuation has to be defended by what comes next, not by what already exists.
For GCF FramWorks LTD, that means moving from a founder-built, pre-revenue AI system into something much closer to an investable technical and commercial asset. The funding plan is roughly split across four areas.
Around 40% of the round goes toward core engineering and founder runway. That’s £100,000. This gives me time to keep building without the company being constrained by my personal finances. A large part of this project has been founder-funded so far, and there is a point where trying to finance serious AI development personally stops being sensible.
Around 20% goes toward compute, model R&D, and training infrastructure. That’s £50,000. ISAC is deliberately built around the idea that useful AI does not always require the largest possible model. One of the major research directions is what I call smart-per-GB, which is roughly how much useful cognition can I get for the amount of compute being used. That means benchmarking smaller models, model layouts, routing, specialist models, fine-tuning, and eventually ISAC-specific models. It also covers things like local runtime work, managed GGUF acquisition, and the GCF resource envelope experiments that are currently in progress. That takes compute.
Around 15% goes toward security, testing, and independent review. That’s £37,500. This one matters a lot to me. If I am going to build an AI system that can eventually use memory, tools, goals, external providers, and real-world capabilities, I do not want the only person testing the architecture to be the person who designed it. The project needs adversarial testing, independent review, security work, claims registers, and proper validation. Some of that work is already in progress. Most of it isn’t.
Around 25% goes toward productisation, pilots, legal, IP, and operating costs. That’s £62,500. There is a massive difference between software that works on the founder’s machine and software that other people can actually use. Installation. Documentation. User experience. Licensing. IP work. Company operations. Customer validation. Potential pilots. All of that starts becoming relevant as ISAC moves toward an actual product. The beachhead target is developers and AI teams who already integrate models and agents, not consumer ISAC, which comes later.
what’s already in the data room
Something I want to be clear about, because I think it matters for the people reading this who might be considering whether to take the conversation further.
The due diligence materials are already prepared. Business plan and financial model. Technical roadmap and stage evidence. Risk, security, and claims registers. Architecture documentation and the procurement data room around the Sovereign AI R&D EOI submission. That is not a future deliverable, it is what is sitting ready now.
I would rather know early if a potential investor sees something in those materials that needs more work than find out six weeks into a conversation.
What £250k should achieve
If I take outside investment, I want to be able to point to a very clear reason for taking it.
The aim would be roughly 12 to 18 months of runway to move the company through a much more valuable set of milestones. That includes things like:
Completing and demonstrating the stable ISAC foundation. Continuing through the remaining roadmap toward ISAC Core v1. Producing proper benchmark evidence. Adding external and custom model support. Building encrypted memory and persistent identity. Adding goals, skills and governed initiative. Testing offline capability. Running adversarial and failure testing. Getting independent technical and security review. Protecting the company’s IP. Beginning proper external validation. And ultimately getting the product into the hands of early users.
That is the justification for the round. The investment is meant to increase the value of the thing being built.
If £250,000 takes GCF FramWorks LTD from a founder-funded technical prototype to a validated AI architecture, a working product, a benchmarked system, protected IP, and something approaching early commercial adoption, then the valuation discussion becomes much easier to understand.
it’s still a risk
I want to say that openly, because I do not think founders should pretend otherwise.
It is still a risk. This is pre-seed. There are things that can fail. There are assumptions that still need testing. There are no paying customers today. I would rather say that openly than pretend otherwise.
But investors are not being asked to value what exists only today. They are being asked to judge whether the current foundation, the architecture, the roadmap, and the team can turn that £250,000 into something substantially more valuable.
That is what they are pricing.
But first I need to finish breaking it
Before any of that, I have Phase 2 to finish. Then the source-control closeout. Then Phase 3. Then another closeout. Then the full post-fix Foundation revalidation. Only after that am I willing to call it Foundation Stable.
Then I want the benchmarks. And then, finally, the live demo.
That one is going to mean a lot to me. Because there is a huge difference between saying “I am building this” and being able to say “here it is. Watch it run.”
Then we move to Stage 36. And things start getting considerably more interesting.
— Billy P., Founder, GCF FramWorks LTD
This is the fourth post in Founder Notes. The first covered the formation of GCF FramWorks LTD. The second covered the Stage 35 finish and the start of the audit. The third covered why GCF is not just another AI agent. This one covers why the audit has to finish before the next round of work begins, and what the £250k pre-seed actually pays for.
Tested on the GCF 2.0 frozen framework, SHA eeae2bc8a426b063c687e40f9724adbb1dc3a61a. The foundation is almost there. Then we can build properly on top of it.