GLOBAL WARMING A system does not become safe because its report says safe. The observable record has to support the claim. I’ve been working on an observable-only public-good review frame for climate-adjacent risk: not accusation, not private-data access, not causation claims. Just a disciplined way to ask whether public burden is showing sustained divergent drift from a claimed stable baseline. Results coming soon...

drive.google.com/file/d/1Mrm… This is a public-good review note for asking whether public data supports a claim of “safe,” “sustainable,” or “under control,” without accusing anyone or needing private information. Use it as a checklist: define the claim, pick public signals, test the baseline, check for drift, run controls, and route only a candidate finding for human review. @NASAClimate @NOAAClimate @IPCC_CH @WRIClimate @CarbonBubble @RockyMtnInst @climate @CDP @CeresNews @grantham_ic @PIK_Climate @FutureEarth @OurWorldInData @ClimateCentral @InsideClimate @SEIresearch

Some organizations have reached out... If you’re wondering whether the geometry I’ve been working on is “real enough,” here’s the answer: when you couple GR U(1) to standard curvature terms and add a covariant scalar layer encoding ΔΔF, you get a fully EFT-safe framework that stays second-order, ghost-free, and hyperbolic. GR lightcones remain intact. Maxwell equations get a tiny Φ-dependent correction. And the warp-shell metric sits inside the solution space without violating known bounds. In other words: the math stands on its own, and it passed several external technical sniff tests.

@METR_Evals, @ApolloAISafety, @PromptFoo, @LangChainAI, @ArizeAI, @GalileoAI, @Langfuse, @HuggingFace, @Scale_AI, @EleutherAI, @AISafetyInst, @DanHendrycks, @CenterForAISafety, @alignmentforum, @DeepEval, @BraintrustData, @WeaveByWandB, @TruEraAI, @Comet_ML, @ContextualAI, @ApolloResearch, @METR_evals, @FAR_AI, @ALIGN_RC, @CAIS_safety, @huggingface, @ScaleAI, @weights_biases, @GovAIOxford, @BAIROpenSource, @StanfordHAI, @MIT_CSAIL, @GoogleDeepMind, @OpenAI, @AnthropicAI, @xai, @MetaAI, @ApolloResearchAI, @RedwoodResearch, @FARAI_org, @Cohere, @MistralAI, @StabilityAI, @CRFM_Stanford

New England should be paying attention to USGS.gov. The northern Appalachians are not just scenery. USGS has flagged regional potential for battery and critical minerals, including lithium, graphite, cobalt, manganese, rare earth elements, tantalum, cesium, tin, beryllium, and niobium. Handled carefully, this could become for parts of New England what oil became for Alaska: not a magic jackpot, but a long-term public wealth, infrastructure, jobs, and supply-chain opportunity. The key is doing it right: map first, protect water, respect towns, build refining capacity, and make sure local people benefit before outside capital extracts the upside. @GovPhilScott @GovJanetMills @KellyAyotte @MassGovernor @GovNedLamont @GovDanMcKee @vtdigger @PressHerald @NHPR @BostonGlobe @CTMirror @projo @wcax @newscentermaine @WMUR9 @WCVB @NBCConnecticut @wpri12 This is a regional strategy conversation now. Your budget problems could be over.

To be clear, I’m not offering a story about UFOs, disclosure, or secret programs. I also don't doubt people who have those stories but I don't have a background in that. I’m offering a civilian black-box AI reliability diagnostic. The problem is simple: as AI systems move into higher-stakes environments, performance is not enough. We need to know whether success comes from real adaptive correction, or from replay, spoofing, brittle mimicry, reward-hacking, or frozen behavior. ΔΔF looks at time-ordered behavior as an observable trajectory. No model weights. No source code. No training data. No proprietary prompts. No classified systems. Just the curve of correction. That matters for AI safety, infrastructure resilience, critical-systems audit, defensive cyber assurance, emergency response, and institutional risk reduction. The boundary is equally simple: Civilian reliability work is in scope. Weapons, targeting, lethal autonomy, battlefield deployment, offensive cyber, military operational optimization, unrestricted sublicensing, derivative military adaptation, and rights-violating surveillance are not. This is not a defense product looking for a loophole. It is a civilian structural-integrity tool for the AI systems society is about to rely on.