Hi all,
I want to share something I’ve been building solo for a while. I’m not a CS grad — my background is Business management — so I don’t have fancy jargon, just a lot of stubbornness and curiosity.
Limitation is your own imagination.
I’m building an AI system from zero: pure C++, no external dependencies, no graph, no static nodes, no vector embeddings, no RAG, no lemmatization, no dictionary, no pre-training, no tokens. Everything hand-built, layer by layer.
The project has 3 parts:
- Neural network layer — done. Works as a sensory/transport layer (mimicking biocomputing signal transmission), not where the “thinking” happens.
- Knowledge + thinking layer — ~65%, still in progress.
- Custom DB — later, when I have the budget
I just finished stress-testing the ingestion daemon (Phase 1) across 3 different environments: my old laptop on Windows, WSL2, and a free Codespace VM. Same < 1MB binary, same test scenario (800M frames, simulated upstream collapse with partial recovery).
The funniest result: on my Windows laptop, I only had ~970MB of free RAM left (out of 8GB) while running the test — browser + 5 Cursor instances open at once. The daemon still passed every gate, no crashes, no data loss, but latency clearly “throttled” it self under pressure:
-
Windows (970MB free RAM): p99 ~16.4µs
-
WSL2 (3.3GB free RAM): p99 ~4.0µs
-
Azure VM (5.7GB free RAM): p99 ~1.3µs
daily_report_2026-07-06_codespace.md.pdf (147.2 KB)
daily_report_20260706_1453_windows.md.pdf (198.0 KB)
That gap made me realize the system was doing exactly what I designed it to do: auto-adapt to available resources instead of forcing through and crashing. Graceful degradation, not failure.
Anyway — none of this would exist without guidance from God, and without Cursor. I’m not a professional engineer, just someone with an idea and a lot of late nights, and Cursor not only tool its my brain extension that let me actually build it instead of just dreaming about it.
Still a long way to go (Phase 2 is only 65% done), but wanted to share this milestone.
Thank you.