https://blog.redwoodresearch.org/archivedailyhttps://blog.redwoodresearch.org/aboutweeklyhttps://blog.redwoodresearch.org/podcastweeklyhttps://blog.redwoodresearch.org/p/blocking-live-failures-with-synchronous2026-03-30monthlyhttps://blog.redwoodresearch.org/p/reward-seekers-will-probably-behave2026-03-28monthlyhttps://blog.redwoodresearch.org/p/ais-capability-improvements-havent2026-03-27monthlyhttps://blog.redwoodresearch.org/p/are-ais-more-likely-to-pursue-on2026-03-12monthlyhttps://blog.redwoodresearch.org/p/the-case-for-satiating-cheaply-satisfied2026-03-10monthlyhttps://blog.redwoodresearch.org/p/frontier-ai-companies-probably-cant2026-02-26monthlyhttps://blog.redwoodresearch.org/p/announcing-controlconf-20262026-02-26monthlyhttps://blog.redwoodresearch.org/p/will-reward-seekers-respond-to-distant2026-02-16monthlyhttps://blog.redwoodresearch.org/p/how-do-we-more-safely-defer-to-ais2026-03-12monthlyhttps://blog.redwoodresearch.org/p/distinguish-between-inference-scaling2026-02-11monthlyhttps://blog.redwoodresearch.org/p/fitness-seekers-generalizing-the2026-01-30monthlyhttps://blog.redwoodresearch.org/p/the-inaugural-redwood-research-podcast2026-01-06monthlyhttps://blog.redwoodresearch.org/p/recent-llms-can-do-2-hop-and-3-hop2026-01-02monthlyhttps://blog.redwoodresearch.org/p/measuring-no-cot-math-time-horizon2025-12-29monthlyhttps://blog.redwoodresearch.org/p/recent-llms-can-use-filler-tokens2025-12-25monthlyhttps://blog.redwoodresearch.org/p/basharena-and-control-setting-design2025-12-18monthlyhttps://blog.redwoodresearch.org/p/the-behavioral-selection-model-for2025-12-04monthlyhttps://blog.redwoodresearch.org/p/will-ai-systems-drift-into-misalignment2025-11-15monthlyhttps://blog.redwoodresearch.org/p/whats-up-with-anthropic-predicting2025-12-25monthlyhttps://blog.redwoodresearch.org/p/sonnet-45s-eval-gaming-seriously2025-10-30monthlyhttps://blog.redwoodresearch.org/p/should-ai-developers-remove-discussion2025-10-23monthlyhttps://blog.redwoodresearch.org/p/is-90-of-code-at-anthropic-being2025-10-22monthlyhttps://blog.redwoodresearch.org/p/reducing-risk-from-scheming-by-studying2025-10-16monthlyhttps://blog.redwoodresearch.org/p/iterated-scheming-testbed-development2025-10-10monthlyhttps://blog.redwoodresearch.org/p/the-thinking-machines-tinker-api2025-10-09monthlyhttps://blog.redwoodresearch.org/p/plans-a-b-c-and-d-for-misalignment2025-10-08monthlyhttps://blog.redwoodresearch.org/p/notes-on-fatalities-from-ai-takeover2025-09-24monthlyhttps://blog.redwoodresearch.org/p/focus-transparency-on-risk-reports2025-09-22monthlyhttps://blog.redwoodresearch.org/p/prospects-for-studying-actual-schemers2025-09-19monthlyhttps://blog.redwoodresearch.org/p/what-training-data-should-developers2025-09-17monthlyhttps://blog.redwoodresearch.org/p/ais-will-greatly-change-engineering2025-09-09monthlyhttps://blog.redwoodresearch.org/p/trust-me-bro-just-one-more-rl-scale2025-09-03monthlyhttps://blog.redwoodresearch.org/p/attaching-requirements-to-model-releases2025-08-27monthlyhttps://blog.redwoodresearch.org/p/notes-on-cooperating-with-unaligned2025-08-24monthlyhttps://blog.redwoodresearch.org/p/being-honest-with-ais2025-08-21monthlyhttps://blog.redwoodresearch.org/p/my-agi-timeline-updates-from-gpt2025-08-20monthlyhttps://blog.redwoodresearch.org/p/four-places-where-you-can-put-llm2025-08-09monthlyhttps://blog.redwoodresearch.org/p/should-we-update-against-seeing-relatively2025-07-28monthlyhttps://blog.redwoodresearch.org/p/why-its-hard-to-make-settings-for2025-07-18monthlyhttps://blog.redwoodresearch.org/p/recent-redwood-research-project-proposals2025-07-14monthlyhttps://blog.redwoodresearch.org/p/guide2025-07-28monthlyhttps://blog.redwoodresearch.org/p/whats-worse-spies-or-schemers2025-07-09monthlyhttps://blog.redwoodresearch.org/p/ryan-on-the-80000-hours-podcast2025-07-08monthlyhttps://blog.redwoodresearch.org/p/how-much-novel-security-critical2025-07-06monthlyhttps://blog.redwoodresearch.org/p/two-proposed-projects-on-abstract2025-07-04monthlyhttps://blog.redwoodresearch.org/p/there-are-two-fundamentally-different2025-07-02monthlyhttps://blog.redwoodresearch.org/p/jankily-controlling-superintelligence2025-06-27monthlyhttps://blog.redwoodresearch.org/p/what-does-10x-ing-effective-compute2025-06-24monthlyhttps://blog.redwoodresearch.org/p/comparing-risk-from-internally-deployed2025-06-23monthlyhttps://blog.redwoodresearch.org/p/making-deals-with-early-schemers2025-08-22monthlyhttps://blog.redwoodresearch.org/p/prefix-cache-untrusted-monitors-a2025-06-21monthlyhttps://blog.redwoodresearch.org/p/ai-safety-techniques-leveraging-distillation2025-06-19monthlyhttps://blog.redwoodresearch.org/p/when-does-training-a-model-change2025-06-12monthlyhttps://blog.redwoodresearch.org/p/the-case-for-countermeasures-to-memetic2025-06-21monthlyhttps://blog.redwoodresearch.org/p/ais-at-the-current-capability-level2025-05-12monthlyhttps://blog.redwoodresearch.org/p/misalignment-and-strategic-underperformance2025-05-08monthlyhttps://blog.redwoodresearch.org/p/training-time-schemers-vs-behavioral2025-05-16monthlyhttps://blog.redwoodresearch.org/p/whats-going-on-with-ai-progress-and2025-05-08monthlyhttps://blog.redwoodresearch.org/p/how-can-we-solve-diffuse-threats2025-04-30monthlyhttps://blog.redwoodresearch.org/p/7-tractable-directions-in-ai-control2025-05-08monthlyhttps://blog.redwoodresearch.org/p/clarifying-ai-r-and-d-threat-models2025-04-25monthlyhttps://blog.redwoodresearch.org/p/how-training-gamers-might-function2025-04-26monthlyhttps://blog.redwoodresearch.org/p/handling-schemers-if-shutdown-is2025-06-21monthlyhttps://blog.redwoodresearch.org/p/ctrl-z-controlling-ai-agents-via2025-04-16monthlyhttps://blog.redwoodresearch.org/p/to-be-legible-evidence-of-misalignment2025-04-24monthlyhttps://blog.redwoodresearch.org/p/why-do-misalignment-risks-increase2025-04-11monthlyhttps://blog.redwoodresearch.org/p/an-overview-of-areas-of-control-work2025-04-09monthlyhttps://blog.redwoodresearch.org/p/an-overview-of-control-measures2025-04-14monthlyhttps://blog.redwoodresearch.org/p/buck-on-the-80000-hours-podcast2025-04-05monthlyhttps://blog.redwoodresearch.org/p/notes-on-countermeasures-for-exploration2025-04-04monthlyhttps://blog.redwoodresearch.org/p/notes-on-handling-non-concentrated2025-04-06monthlyhttps://blog.redwoodresearch.org/p/prioritizing-threats-for-ai-control2025-03-19monthlyhttps://blog.redwoodresearch.org/p/how-might-we-safely-pass-the-buck2025-02-19monthlyhttps://blog.redwoodresearch.org/p/takeaways-from-sketching-a-control2025-01-30monthlyhttps://blog.redwoodresearch.org/p/planning-for-extreme-ai-risks2025-01-30monthlyhttps://blog.redwoodresearch.org/p/ten-people-on-the-inside2025-01-28monthlyhttps://blog.redwoodresearch.org/p/when-does-capability-elicitation2025-01-22monthlyhttps://blog.redwoodresearch.org/p/how-will-we-update-about-scheming2025-05-20monthlyhttps://blog.redwoodresearch.org/p/thoughts-on-the-conservative-assumptions2025-01-17monthlyhttps://blog.redwoodresearch.org/p/extending-control-evaluations-to2025-02-12monthlyhttps://blog.redwoodresearch.org/p/measuring-whether-ais-can-statelessly2025-02-12monthlyhttps://blog.redwoodresearch.org/p/alignment-faking-in-large-language2024-12-18monthlyhttps://blog.redwoodresearch.org/p/why-imperfect-adversarial-robustness2024-12-17monthlyhttps://blog.redwoodresearch.org/p/wincontinuelose-scenarios-and-executereplaceaudi2025-01-14monthlyhttps://blog.redwoodresearch.org/p/behavioral-red-teaming-is-unlikely2024-10-10monthlyhttps://blog.redwoodresearch.org/p/a-basic-systems-architecture-for2024-09-26monthlyhttps://blog.redwoodresearch.org/p/how-to-prevent-collusion-when-using2024-09-25monthlyhttps://blog.redwoodresearch.org/p/would-catching-your-ais-trying-to2024-08-26monthlyhttps://blog.redwoodresearch.org/p/fields-that-i-reference-when-thinking2024-08-13monthlyhttps://blog.redwoodresearch.org/p/getting-50-sota-on-arc-agi-with-gpt2025-06-08monthlyhttps://blog.redwoodresearch.org/p/access-to-powerful-ai-might-make2025-06-08monthlyhttps://blog.redwoodresearch.org/p/ai-catastrophes-and-rogue-deployments2024-06-03monthlyhttps://blog.redwoodresearch.org/p/preventing-model-exfiltration-with2024-05-08monthlyhttps://blog.redwoodresearch.org/p/untrusted-smart-models-and-trusted2024-05-07monthlyhttps://blog.redwoodresearch.org/p/managing-catastrophic-misuse-without2024-05-07monthlyhttps://blog.redwoodresearch.org/p/catching-ais-red-handed2024-05-08monthlyhttps://blog.redwoodresearch.org/p/the-case-for-ensuring-that-powerful2024-12-05monthly