A home for some of my work on models, meaning, simulation, privacy, governance & human-AI systems. Six essays/preprints, mirrored on GitHub with their canonical Zenodo DOIs preserved. github.com/putmanmodel/model…

For anyone looking to save some time on going to church, this basically sums it all up. All hail Garth Marenghi. youtube.com/shorts/IBQh5ejFh…

While integrating CDE-Kingpin with NVIDIA OpenShell, operation binding surfaced an upstream bug: plaintext HTTP was reported to middleware as HTTPS. I stopped, reported #4253, it was independently reproduced, fixed in #4397, and merged. Candidate A + B now pass. github.com/putmanmodel/agent…

Unity devs: Looking for someone making a game with NPCs to privately test Spanda. It adds context- and experience-aware behavior to existing NPC controllers, with INSPECT to explain decisions. No LLM required. Codex/Claude Code setup guides included. Interested? DM me. #unity3d

Unity developers: Can I borrow an NPC? I'm developing Spanda, an experimental system that lets NPCs interpret interactions differently based on context, relationships, and previous experiences. The goal isn't to replace your NPC controller or dialogue system, but to work alongside what you've already built. I'm looking for a developer willing to let me experiment with integrating Spanda into an existing NPC. Nothing needs to be fancy. A basic NPC with a dialogue system or a few callable behaviors would be perfect. This isn't a finished SDK or a product pitch. I'm looking to test interoperability with real developer projects. Interested? Reply or DM.

Can NPCs learn from experience without gaining unrestricted authority to act? Spanda: Qwen interpretation, governed movement, and bounded adaptive learning in Unity. 🎬 youtu.be/yZFEq3CHnKY 📖 github.com/putmanmodel/spand… #GameAI #Unity

Had to pause the OpenShell-specific CDE/Kingpin supervisor-middleware work after hitting an HTTP/HTTPS scheme bug in the OpenShell path. A kind contributor picked it up almost immediately and put in a vouch request to fix it, but the upstream change is still waiting on triage. Rather than leave CDE/Kingpin blocked on another project’s timeline, I’m going to move forward by making the enforcement boundary substrate-neutral. OpenShell can remain one adapter once the issue is resolved, while the core authority/binding work continues independently. Pause the integration, not the project.

East of Eden, if not my favorite novel outright, is definitely my favorite Steinbeck (and I’ve read most of his work). I’m not sure about the Netflix series, though. If it’s awful, I’ll have to scrub it from my memory by rereading the novel.

More concern. Not more authority. I built a deterministic weak-signal demo around persistence, independent corroboration, provenance, contradiction, quarantine, evidence expiry, and human review before consequential action. Demo: putmanmodel.github.io/weak-s… Repo: github.com/putmanmodel/weak-…

I hope the Cornell victim has better luck with NY AG Letitia James @NewYorkStateAG than I did. This is the response I received from her Civil Rights Bureau after seeking help concerning state and local police conduct. Nothing meaningful has changed since then, and my family continues to experience conduct that I believe warrants outside scrutiny. My concern is that when complaints repeatedly end without meaningful review or accountability, the practical message to local institutions can become that there are few consequences.

I’ve been working on Spanda a lot lately, and I’ve significantly expanded its ability to demonstrate some surprisingly rich behavior deterministically. Systems like this are going to be a big part of what makes the next generation of games and simulations feel dramatically more alive, adaptive, and believable.

What happens to an agent’s authority after approval, revocation, delegation, restart, or uncertain execution? I built the PUTMAN Agent Governance Test Harness to test those transitions against observable runtime evidence. github.com/putmanmodel/agent…

Interesting to see NVIDIA OpenShell arrive around agent runtime safety and enforcement. I’ve been building CDE/Kingpin around stateful runtime authority — contraction, revocation, review, recovery and execution uncertainty. I’m particularly interested in whether Kingpin could complement something like OpenShell as the authority layer while OpenShell provides the stronger enforcement substrate. That seems worth testing.

I wonder how many people I’ve fooled with my expert beard contouring to hide my chubby neck.

What happens to an agent’s authority after approval, revocation, delegation, restart, or uncertain execution? I built the PUTMAN Agent Governance Test Harness to test those transitions against observable runtime evidence. github.com/putmanmodel/agent… #AIAgents #AgentGovernance

I wonder if I’m the only one in #BillsMafia making chickpea curry and naan for the game today.

Question for the local-LLM tech bros: What small-ish local model would you use today for really good instruction adherence + persona-preserving expression? I’m testing models with Spanda, where deterministic NPC relationship state controls the stance and the LLM only expresses it. Gemma 3 4B has been my baseline, but I think I may finally be asking more of it than a 4B model can reliably deliver.