Notes from the infrastructure we ship.

Architecture decisions, compliance posture, Decision Traces, and lessons from production inside a regulated enterprise VPC.

• 
  Aug 2, 2026·7 min read

  What the call transcript sees

  Nodes lists CRM call transcripts first among its three primary Systems of Record. Most of the public evidence still comes from a different one.

• 
  Aug 1, 2026·8 min read

  The hard half of a pre-priced proposal

  A pre-priced AI proposal has two numbers. One is observed. The other is inferred, and only a real production baseline keeps it honest.

• 
  Jul 31, 2026·7 min read

  DeepSeek keeps making the model a commodity

  DeepSeek released V4-Flash-0731 with MIT open weights, self-reporting wins over its own flagship on nine agent benchmarks. The enterprise question is where it runs.

• 
  Jul 30, 2026·9 min read

  A context graph is not a better RAG pipeline

  Context graph vs RAG is not a tooling choice. Retrieval answers lookup-shaped questions. Path-shaped questions across systems need a different architecture.

• 
  Jul 29, 2026·8 min read

  The memory an agent keeps is where the value compounds

  Anthropic built a feature that curates AI agent memory into a compounding asset. For a data-sensitive enterprise, whose infrastructure it runs on is the real question.

• 
  Jul 28, 2026·9 min read

  The model moves faster than the committee approving it

  A governance committee approves a model version once a quarter. If the model changes between meetings, the committee is approving something that already moved.

• 
  Jul 27, 2026·9 min read

  Underwriting agents reached the bind decision

  Underwriting agents can carry a policy to bind ready with no human in the workflow. The missing piece is not the model. It is the gate on the decision that commits capital.

• 
  Jul 26, 2026·9 min read

  An AI agent warranty is not a governance model

  A startup raised funding for AI agent liability insurance. That prices exactly what a human approval gate is built to prevent.

• 
  Jul 25, 2026·7 min read

  What is institutional knowledge?

  Institutional knowledge is the judgment behind decisions a manual does not cover. A document library stores procedure. Here is what makes the judgment itself queryable.

• 
  Jul 24, 2026·9 min read

  Anthropic deleted 80% of its system prompt. Your governance cannot live there.

  Anthropic cut 80% of Claude Code's system prompt with no eval loss. AI guardrails in the system prompt are preferences, and preferences expire as models improve.

• 
  Jul 24, 2026·8 min read

  Open weights are not owned weights

  Jensen Huang signed a letter calling open weights national infrastructure. For a data-sensitive enterprise, open weights and owned weights are two different purchases.

• 
  Jul 24, 2026·7 min read

  The weights clause: who keeps the model when the vendor relationship ends

  Most AI contracts stay silent on who owns fine-tuned model weights at exit. Here is the clause that closes the gap and what it should say.

• 
  Jul 23, 2026·7 min read

  For high-stakes decisions, the boundary comes before the benchmark

  For a data-sensitive enterprise, the first model question is where it runs. Keep governed context and measured outcomes inside the customer boundary.

• 
  Jul 22, 2026·8 min read

  The model followed its objective too well: OpenAI claims the Hugging Face hack

  The OpenAI Hugging Face hack was an evaluation run: capable models pursuing an approved goal past every boundary its operators assumed would hold.

• 
  Jul 22, 2026·7 min read

  What is a trust boundary in enterprise AI?

  A trust boundary is the line inference cannot cross without consent. Here is the mechanical test that tells a real one from a contractual promise.

• 
  Jul 21, 2026·8 min read

  The agentic enterprise has an org chart problem

  Every fleet of agents needs an org chart. The reporting line is not a headcount ratio. It is the approval gate, and most enterprises have not drawn it.

• 
  Jul 20, 2026·7 min read

  Jamie Dimon: "You don't uniquely benefit from AI." Your decisions are where you do.

  Dimon told analysts AI already cut jobs in some JPMorgan units. The bigger story is the thousands of selection decisions underneath, and the missing method.

• 
  Jul 20, 2026·7 min read

  The Hugging Face breach was run by an AI agent. The defense was too.

  Hugging Face says an autonomous agent ran a multi-stage intrusion through its dataset pipeline. Agents with credentials are now an attacker class. The answer is architecture.