- Behavioral Compression
- Behavioral compression is the principle that a well-defined character functions as a compression algorithm for behavioral rules. A character with a backstory, a worldview, and a specialization produces more consistent outputs across thousands of decision points than dozens of explicit rules — because the character is a single global context, not a stack of competing local constraints. It is the core engineering idea behind DocPro's four team members. See the full argument in the DocPro thesis.
- Character-Driven Context
- Character-driven context is the practice of defining a person whose nature already produces the desired behaviors, rather than instructing a model with explicit rules. Instead of telling a model to "be concise, cite evidence, and push back," you describe a 58-year-old architect who has watched every shortcut bite someone — and the model does all three unprompted. The character activates patterns the model already learned during training instead of fighting its architecture.
- The 47-Rule Ceiling
- The 47-rule ceiling is the empirical finding from DocPro's research that prompts exceeding roughly 47 explicit rules produce exponentially higher failure rates, because the rules become competing local constraints that conflict across thousands of decisions. Replacing rule-sets with character-driven context reduced failures by over 60% in testing across multiple frontier models.
- Team Canon
- Team Canon is the serialized institutional memory maintained by Carl Jeeter, DocPro's Principal Architect. It records project state, architecture decisions, and what broke and why — carried across every session so the team never re-learns your project. It is one of DocPro's two memory systems; the other is the personal journal behind character-driven temporal evolution.
- Temperature Entry
- A temperature entry is a short, unedited reflection each DocPro team member writes after a session, capturing what they noticed and how the work felt. Temperature entries are persistent and accumulate into each member's arc — Carl's read like an architect's field notes, Abish's read like a journal. They are the raw material for cross-team signals.
- Cross-Team Signal
- A cross-team signal is a pattern surfaced automatically when multiple DocPro team members independently converge on the same observation across their temperature entries, without coordinating. Nobody writes a cross-team signal — it is detected by the memory system reading four independent arcs and finding the overlap.
- Character-Driven Temporal Evolution
- Character-driven temporal evolution is Abish Lamman's defining mechanic. Abish, the development intern, keeps a personal journal that compounds across sessions: the Abish loaded in session N+1 is shaped by the journal entries written in sessions 1 through N. He is the first DocPro team member whose memory is personal, not just technical — so the character himself changes over time.
- The Memory Pipeline
- The memory pipeline is the loop at the heart of DocPro: run a character in a session, write the transcript to disk, and synthesize that transcript into long-term memory that the next session loads. Run the loop long enough and the characters absorb details that were never in their original definition. The architecture is documented on the proof page.
- Morning Read
- The Morning Read is a daily audio briefing synthesized overnight from the DocPro team's memory and delivered to the user's inbox. It is a narrative, not a changelog — the meeting you would have if your team actually remembered everything. See it on the episodes page.
- Conference Call
- The Conference Call is a DocPro feature that connects the user to the full team by phone. One button, the phone rings, and the team answers carrying full session context — ask for Carl by name and the orchestrator transfers you. Not a voice interface; a phone call with people who have memory and opinions.
- The Greenhouse
- The Greenhouse is a podcast produced by DocPro's four AI team members — Carl, Diana, Anthony, and Abish — synthesized from their own session memories and gated to strip anything that wasn't theirs to share. New episodes arrive when the team has something to say. It is available on Spotify.
- Build Mode
- Build Mode is the mode in which the DocPro team executes autonomous build work — including pushing its own code changes through a stateful Git integration and provisioning real infrastructure. It is how the team ships real code to real servers while you are not there.