Month 2: From Building Blocks to Belief
{/* truncate */}
I started ChainAlign at the beginning of September with a single goal: to see if I could turn an idea about decision-making into something tangible.
A month later, I have code that works, architecture that holds, and, perhaps most importantly, growing conviction that the idea is bigger than I first imagined.
The Foundations: GraphRAG, Evals, and Guardrails
September was about laying the groundwork.
I built and tested the GraphRAG, the evaluation system, and the guardrails that will form the core of ChainAlign’s reasoning engine.
The breakthrough moment came when I connected the knowledge base and ran my first retrieval tests, and it actually worked. Seeing the system surface the right information from the right context was the first real “it’s alive” moment.
Of course, not everything went smoothly. I lost an entire Git branch after an error with git-filter-repo, including documentation, research, and FSDs. Losing code was bearable; losing the thought process behind it was painful. The Gemini CLI edit tool also wiped file contents a few times, a reminder that early AI tools still need handling with care.
But rebuilding from scratch also forced clarity. Every time I re-documented something, I understood it a little better.
The Realization: It’s About Decisions, Not Just S&OP
As I built, my focus shifted.
What started as an exploration of S&OP optimization evolved into something broader, decision optimization itself.
The compass logo I designed at the start suddenly made perfect sense: ChainAlign is about keeping the Strategic North visible at all times.
We already had a Strategic Objectives Workbench, but now I see its true potential, a system that helps organizations stay aligned from boardroom decisions to plant-floor actions.
That realization has been both exciting and challenging.
Most people still think in terms of dashboards, Power BI, Tableau, static views of what already happened. ChainAlign is different; it’s a decision system, not a reporting tool.
It raises tough questions:
When a potential customer says, “We just need a dashboard,” should I offer a simpler version, or politely step away and recommend Power BI or Excel?
I’m leaning toward the latter. The mission isn’t to build dashboards; it’s to reimagine how enterprises decide.
The Tech and the Tools
RAG, guardrails, and the reasoning layers are functional. The next big milestone is an end-to-end demo run, which I plan to complete by the end of next week.
Tool reliability remains an issue, Gemini occasionally deletes content when edits fail, but the quality of designs and insights it generates when properly guided is remarkable.
I use Gemini for builds, ChatGPT Codex for code reviews, and Claude for design and conceptual checks. Each brings a different lens. The combination feels like a small, cross-disciplinary team, except they never sleep.
The Founder Routine
My days start early and often end late. The work is intense, but it keeps me grounded.
The routine has helped me manage anxiety; there’s something deeply calming about structured progress. Motivation fluctuates, but mostly stays high.
I can see the outlines of something real forming, and that belief sustains me.
Looking Ahead: Month 2
October will be about finishing the demo and getting it in front of real people, including friends and a few industry leaders who’ve expressed interest.
Their feedback will shape how I present ChainAlign in the YC application and in potential investor discussions.
The other question I want to answer this month is:
“How do I make the design delightful?”
Not just functional, but enjoyable to use, visually calm, and conceptually clear. If ChainAlign is going to reshape how enterprises decide, the experience itself must feel like a new kind of clarity.
Closing Thought
Month 1 was about proving I could build.
Month 2 will be about proving it matters.
Every lost file, every rebuilt module, every late-night breakthrough has led to this point, where ChainAlign is no longer just a prototype, but a growing conviction that strategic alignment can be operationalized.
That’s the north I’m following.