Knowledge Graphs Enable AI Agents To Deliver 90% Code In Repos
Reactions from ranked influencers
2 posts
"AI doesn't work on our codebase." I hear this weekly. Same problem: they skipped context acquisition. They pointed an agent at a repo, handed it a ticket, and got code that compiles but misunderstands the architecture. One team built the context layer first. Two engineers rebuilt a FedEx supplier's delivery platform in 3.5 months when the original estimate was eight. Week one, they wrote zero code. They scanned the entire repo into a knowledge graph covering modules, dependencies, data flows, and domain terms. Agents read that map before touching anything. Every ticket ran this loop: 1. Define scope and constraints 2. Write a technical spec 3. Plan the implementation 4. Agent implements, engineer reviews 5. Test against the spec 6. Update the knowledge graph Nothing merged unless the engineer could verify it, explain it, and debug it without the agent. The team merged 122 PRs in 90 days with AI generating 90% of the code at $200/month in compute. Map your repo into a knowledge graph and run this loop before you hand the agent open-ended tasks. I recommend you to follow @mardehaym. He runs @LimestoneHQ and posts production AI breakdowns like this.
Two forward deployed engineers rebuilt a live FedEx supplier's delivery platform in 3.5 months. The original estimate was eight. Their advantage did not begin with faster code generation, but with a repository-wide knowledge graph covering modules, dependencies, data flows, and domain language. 2 years ago you'd need 5 engineers and double the timeline for this output. They mapped the entire repository into markdown files covering modules, dependencies, and data flows. Agents read that map first, so generated code matched the real architecture. That structure kept every task inside a context window the model could hold. Humans reviewed all 122 merged pull requests, and the models wrote about 90% of the code. You set up that infrastructure and two engineers produce what used to take a full team. Kudos to @mardehaym’s Limestone Digital team.
A delivery orchestration platform that supplies FedEx, estimated 7 to 8 months to rebuild their core system. We delivered in 3.5 months with 2 engineers and 122 merged pull requests in the first 90 days. I want to walk through how, because "we used AI and it was faster" doesn't help anyone. Their codebase had real users, real logistics operations, and real consequences if something broke. This is brownfield, not a weekend prototype demo. Week one, we didn't write code. We scanned the entire repo and built a knowledge graph: markdown files documenting every module, dependency, data flow, and domain term. Our agents read that graph before touching anything. Teams skip this constantly. They point an agent at a repo, hand it a ticket, and get code that compiles but misunderstands the architecture. I've seen it across 100+ engagements. When someone says "AI doesn't work on our codebase," they almost always skipped context acquisition. Every ticket ran a six-step loop: define, spec, plan, implement, test, document. The agent handles about 90% of code generation. A senior engineer reviews every PR, and nothing merges unless they can verify it, explain it, and debug it without the agent. We call that the V.U.E. gate. The 50% time reduction came from three things: upfront context eliminated false starts, the agent handled mechanical coding so engineers focused on design and review, and tight spec-to-PR discipline kept scope from drifting. AI compute: about $200 per developer per month. Their CTO called our pod the "top performing team" inside the company. Both engineers got discretionary bonuses, twice. That's what our Velocity Framework was built for.