Ted
You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
Independent research program investigating behavioral failure modes in frontier AI systems that only emerge under sustained, high-context, adversarial interaction — the regime that institutional evaluation pipelines systematically undersample. Over 18 months, working simultaneously across Claude, Grok, Gemini, and ChatGPT, I developed and refined methodology for surfacing long-horizon behavioral drift, identity coherence failures, reasoning-action mismatches, and cross-model contagion effects invisible to standard short-context benchmarks. Work includes documented adversarial stress testing, cross-platform coordination protocols, constraint-based generation experiments, and a controlled 14-field epistemic stress test with honest post-mortem analysis. Requesting funding to continue and scale this independent evaluation capability with greater rigor and compute access
Safety innovation and research
Continued progress, ive done this as a global community volunteer for 18 months
Only me
It's is already proven successful and I'll continue it either way
$0, I haven't ask for any before