Examples

2026-10-12

PSE R-STDP 3.0.0 beta

How does a brain learn from reward? PSE R-STDP lets you watch it happen on a small spiking network. It runs reward-modulated spike-timing-dependent plasticity experiments from the command line, in phases and seeds you control, and ships a self-test so you can trust the numbers before you trust the conclusions. A compact, reproducible lab for reward-driven learning, numpy and nothing else. Free and open source (GPL-3.0), for Ubuntu and Windows. Release notes

What PSE does here: PSE's Care Invariant is the neuromodulator: the third factor that decides how far back a reward can assign credit.

2026-10-12

PSE-Jacobian 0.1.0 beta

The Jacobian Conjecture stood open for 87 years. In July 2026 Levent Alpöge published a counterexample in three variables, found with Claude Fable 5. The two-variable case is still open. PSE-Jacobian lets you check the counterexample yourself, on your own machine. It reproduces the construction over the complex numbers from scratch, assembles the map and its Jacobian symbolically, and confirms the determinant is a nonzero constant while the map fails to invert, the exact shape of a counterexample. Every run is sealed with a hash and a plain-language report; two Guardians keep it honest and nothing is asserted that the computation does not show. Free and open source (GPL-3.0), for Ubuntu and Windows. Release notes

What PSE does here: Three independent units of PSE agents audit the map, each in its own thread, and a result counts only when all three agree.

2026-10-12

PSE-Kali 0.1.0 beta

A penetration test is only as trustworthy as its restraint. PSE-Kali puts a PSE Guardian between a language model and the real tools, nmap, tcpdump, ssh, hashcat and the rest, so every action is checked, logged and bounded before it runs against an authorized target. One kernel, one shared findings store, a live view of what it is doing and why. Defensive, authorized-assessment use only; the mock backend runs fully offline. Free and open source (GPL-3.0), for Ubuntu. Release notes

What PSE does here: A PSE Guardian stands between the language model and the real tools, and checks, logs and bounds every action before it runs.