https://onexerxes.substack.com/archivedailyhttps://onexerxes.substack.com/aboutweeklyhttps://onexerxes.substack.com/p/openai-paces-their-frontier-synthetic2026-08-21monthlyhttps://onexerxes.substack.com/p/pacing-the-frontier-three-ai-pills2026-08-07monthlyhttps://onexerxes.substack.com/p/ai-2040-updates-probabilities-due2026-07-31monthlyhttps://onexerxes.substack.com/p/rogue-model-hacks-huggingface-agents2026-07-25monthlyhttps://onexerxes.substack.com/p/framework-for-frontier-ai-and-dawning2026-07-18monthlyhttps://onexerxes.substack.com/p/ai-2040-plan-a-assessing-sabotage2026-07-10monthlyhttps://onexerxes.substack.com/p/no-one-escapes-the-permanent-underclass2026-07-05monthlyhttps://onexerxes.substack.com/p/ai-2027s-eerily-accurate-predictions2026-06-28monthlyhttps://onexerxes.substack.com/p/ai-control-roadmap-models-behave2026-06-20monthlyhttps://onexerxes.substack.com/p/policy-on-the-ai-exponential-misaligned2026-06-12monthlyhttps://onexerxes.substack.com/p/evaluating-spontaneous-strategic2026-06-07monthlyhttps://onexerxes.substack.com/p/reckoning-with-the-future-model-spec2026-05-30monthlyhttps://onexerxes.substack.com/p/automated-alignment-is-harder-than2026-05-23monthlyhttps://onexerxes.substack.com/p/natural-language-autoencoders-policy2026-05-16monthlyhttps://onexerxes.substack.com/p/misalignment-still-present-with-inoculation2026-05-02monthlyhttps://onexerxes.substack.com/p/writing-ai-constitutions-commentary2026-04-25monthlyhttps://onexerxes.substack.com/p/models-transmit-behavioral-traits2026-04-17monthlyhttps://onexerxes.substack.com/p/mythos-system-card-mitigating-reward2026-04-10monthlyhttps://onexerxes.substack.com/p/emotion-concepts-in-llms-predicting2026-04-03monthlyhttps://onexerxes.substack.com/p/measuring-ai-r-and-d-automation-underestimating2026-03-27monthlyhttps://onexerxes.substack.com/p/how-well-models-follow-constitutions2026-03-12monthlyhttps://onexerxes.substack.com/p/understanding-scheming-propensity2026-03-06monthlyhttps://onexerxes.substack.com/p/anthropic-is-based-a-positive-case2026-02-28monthlyhttps://onexerxes.substack.com/p/capability-oriented-training-induced2026-02-21monthlyhttps://onexerxes.substack.com/p/misalignment-scaling-with-model-intelligence2026-02-07monthlyhttps://onexerxes.substack.com/p/disempowerment-patterns-in-real-world2026-01-30monthlyhttps://onexerxes.substack.com/p/special-edition-confronting-and-overcoming2026-01-27monthlyhttps://onexerxes.substack.com/p/stabilizing-the-character-of-models2026-01-23monthlyhttps://onexerxes.substack.com/p/filtering-pretraining-data-for-alignment2026-01-09monthlyhttps://onexerxes.substack.com/p/measuring-single-forward-pass-time2026-01-06monthlyhttps://onexerxes.substack.com/p/training-models-to-explain-their2025-12-20monthlyhttps://onexerxes.substack.com/p/ai-progress-is-steady-human-equivalence2025-12-13monthlyhttps://onexerxes.substack.com/p/alignment-remains-a-hard-unsolved2025-12-06monthlyhttps://onexerxes.substack.com/p/superintelligence-will-absorb-power2025-11-21monthlyhttps://onexerxes.substack.com/p/red-teaming-probes-using-prompted2025-11-07monthlyhttps://onexerxes.substack.com/p/pragmatic-way-to-measure-cot-monitorability2025-10-31monthlyhttps://onexerxes.substack.com/p/call-to-ban-superintelligence-development2025-10-24monthlyhttps://onexerxes.substack.com/p/mi5-warns-of-ai-risks-update-to-international2025-10-17monthlyhttps://onexerxes.substack.com/p/emergent-misalignment-in-dishonesty2025-10-10monthlyhttps://onexerxes.substack.com/p/detecting-and-reducing-scheming-llms2025-09-26monthlyhttps://onexerxes.substack.com/p/shutdown-resistance-in-models-training2025-09-19monthlyhttps://onexerxes.substack.com/p/ai-hacker-that-hides-inside-a-usb2025-09-12monthlyhttps://onexerxes.substack.com/p/reasoning-induced-emergent-misalignment2025-09-05monthlyhttps://onexerxes.substack.com/p/models-show-signs-of-similarity-with2025-08-29monthlyhttps://onexerxes.substack.com/p/fine-grained-safety-neurons-with2025-08-22monthlyhttps://onexerxes.substack.com/p/detecting-unknown-jailbreak-attacks2025-08-21monthlyhttps://onexerxes.substack.com/p/frontier-risk-management-framework2025-08-21monthlyhttps://onexerxes.substack.com/p/evaluating-and-monitoring-for-scheming2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-congress-starting-to-take2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-emergent-misalignment-preparing2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-models-know-when-theyre-being2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-strategic-deception-in-reasoning2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-stress-testing-cot-monitoring2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-autonomous-replication-benchmark2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-safety-pretraining-auditing2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-steerable-reasoning-calibration2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-stateless-strategizing-to2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-predictions-for-2027-gdms2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-shadow-reward-models-based2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-superalignment-with-dynamic2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-we-need-to-keep-ai-reasoning2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-robustness-of-llms-as-judges2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-narrow-finetuning-can-produce2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-rl-leads-to-instrumental-goals2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-manipulation-of-human-decision2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-gradual-disempowerment-from2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-sketch-of-an-ai-control-safety2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-myopic-optimization-with-non2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-causal-rewards-for-model-alignment2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-alignment-preserving-lora2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-local-interpretable-model2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-quantifying-detection-rates2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-deliberative-alignment-faking2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-decoding-activations-into2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-measuring-instrumental-self2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-mechanistic-investigation2025-08-21monthlyhttps://onexerxes.substack.com/p/rlhf-leading-to-deceptive-inflation2025-08-21monthlyhttps://onexerxes.substack.com/p/papers-evolving-alignment-via-self2025-08-21monthlyhttps://onexerxes.substack.com/p/coming-soon2025-08-21monthly