Opinion

Deeper Reviews for the top 15 (of the 2024 Review)

​Published on January 14, 2026 11:59 PM GMTWe’re extending the Discussion Phase of the 2024 Annual Review. One thing I’m particularly hoping for is to get more in-depth reviews (especially critical ones) of the posts that currently look likely to be in the top-10 or so. (Ideally the entire top 50, but seemed particularly worthwhile to give serious evaluations of the most heavily upvoted ideas).Eyeballing the reviews and top posts so far, I think posts could use to get more thorough, evaluative attention than they’re currently getting.There’s been some debate this year about whether we should care more about “what’s great about a post” vs “what was lacking.” Ideally, we’d have posts that are great without major flaws, and in my ideal world, the Annual Review results in posts with notable errors getting fixed.In practice, people disagree about what counts as an error, or what errors are particularly egregious. Some errors are quick for an author to fix, some are more gnarly and maybe the author disagrees about the extent to which they are an error.The solution we’ve found for now is to make reviews more prominent on Best Of LessWrong posts, and try to aim for a world where if there is major disagreement, controversy or important considerations about a post, future people will see that disagreement.Currently we do this by including a one-line comment wherever the Spotlight showsup. We may invest more in that over time. This also means if you wrote a review that got 10+ karma, it’s probably worth optimizing the first line to convey whatever information you’d like someone casually skimming the site to read. You can look over the current leaderboard for reviewers to get a sense of which of your reviews might be worth polishing.[1]If you know of someone who’s already written a blogpost or other public response to some works, it’d be helpful to write a short review linking to it and explaining it’s most significant takeaways.The Top 50 as of Jan 1We don’t continuously update the results of the nomination votes, to discourage strategic voting. But, here were the results as-of a couple weeks ago.You might want to note both whether there are posts you think are over/underrated that you want to write in support of. Posts need at least 1 review to make it to the final voting phase.#TitleReviewers1Alignment Faking in Large Language Modelsryan_greenblattJan_KulveitjohnswentworthMarcelo Tibau2On greenJoe CarlsmithRaymond Douglas3Believing InAnnaSalamonBen Pace4The hostile telepaths problemValentineValentineMartin RandallRubyGordon Seidoh WorleyHugo LLucie Philippon5Reliable Sources: The Story of David GerardTracingWoodgrainsThomas KwaScrewtape6Overview of strong human intelligence amplification methodsTsviBTkave7The case for ensuring that powerful AIs are controlledryan_greenblattAlexa Pan8The impossible problem of due processmingyuanAnnaSalamon9Deep atheism and AI riskJoe CarlsmithNO REVIEWS[2]10NeutralitysarahconstantinAnnaSalamon11And All the Shoggoths Merely PlayersZack_M_DavisMartin RandallSeth Herd12Truthseeking is the ground in which other principles growElizabethSherrinford13My Clients, The LiarsymeskhoutScrewtape14Gentleness and the artificial OtherJoe CarlsmithNO REVIEWS[2]15″How could I have thought that faster?”mesaoptimizerNO REVIEWS[2]16Safety isn’t safety without a social model (or: dispelling the myth of per se technical safety)Andrew_CritchAlexa PanZac Hatfield-Doddsthe gears to ascension17What Goes Without SayingsarahconstantinZac Hatfield-Dodds18Repeal the Jones Act of 1920ZviThomas Kwa19Would catching your AIs trying to escape convince AI developers to slow down or undeploy?BuckAlexa Pan20On attunementJoe CarlsmithZac Hatfield-Dodds21There is way too much serendipityMalmesburyGordon Seidoh Worley22The Summoned Heroine’s Prediction Markets Keep Providing Financial Services To The Demon King!abstractapplicScrewtapeabstractapplic23Towards a Broader Conception of Adverse SelectionRicki HeickleneggsyntaxNathan Young24The Field of AI Alignment: A Postmortem, and What To Do About ItjohnswentworthThomas Kwa25Arithmetic is an underrated world-modeling technologydynomightBen Pace26“Alignment Faking” frame is somewhat fakeJan_KulveitJan_Kulveit27Simple versus Short: Higher-order degeneracy and error-correctionDaniel MurfetDaniel MurfetZack_M_Davisniplav28’Empiricism!’ as Anti-EpistemologyEliezer YudkowskyBen Pace29Catching AIs red-handedryan_greenblattBuck30A Three-Layer Model of LLM PsychologyJan_KulveitGunnar_ZarnckeJan_Kulveit31My AI Model Delta Compared To ChristianojohnswentworthRuby32Superbabies: Putting The Pieces TogethersarahconstantinNO REVIEWS[2]33AI catastrophes and rogue deploymentsBuckBuck34Circular Reasoningabramdemskiplex35ThresholdingReview BotScrewtape36Interpreting Quantum Mechanics in Infra-Bayesian PhysicalismYegregVanessa Kosoy37Preventing model exfiltration with upload limitsryan_greenblattNoosphere8938shortest goddamn bayes guide everlemonhopeScrewtape39Transformers Represent Belief State Geometry in their Residual StreamAdam ShaiAdam Shai40Why I’m not a BayesianRichard_NgoRichard_NgoRichard Korzekwa Nathan Young41Hierarchical Agency: A Missing Piece in AI AlignmentJan_KulveitVanessa Kosoy42Being nicer than ClippyJoe CarlsmithNO REVIEWS[2]43You don’t know how bad most things are nor precisely how they’re bad.Solenoid_EntitySolenoid_Entity44Struggling like a ShadowmothRaemonRaemon45Why Don’t We Just… Shoggoth+Face+Paraphraser?Daniel KokotajloDaniel Kokotajlo46The Inner Ring by C. S. LewisSaul MunnNathan Young47Anvil ShortageScrewtapeScrewtapeThomas KwaLorxus48Raising children on the eve of AIjuliawiseNO REVIEWS[2]49Priors and PrejudiceMathiasKBScrewtape50[Intuitive self-models] 6. Awakening / Enlightenment / PNSESteven Byrneslsusr ^(Note, these karma scores subtract your own self-upvote)^Posts without reviews won’t appear in Final Voting PhaseDiscuss ​Read More

​Published on January 14, 2026 11:59 PM GMTWe’re extending the Discussion Phase of the 2024 Annual Review. One thing I’m particularly hoping for is to get more in-depth reviews (especially critical ones) of the posts that currently look likely to be in the top-10 or so. (Ideally the entire top 50, but seemed particularly worthwhile to give serious evaluations of the most heavily upvoted ideas).Eyeballing the reviews and top posts so far, I think posts could use to get more thorough, evaluative attention than they’re currently getting.There’s been some debate this year about whether we should care more about “what’s great about a post” vs “what was lacking.” Ideally, we’d have posts that are great without major flaws, and in my ideal world, the Annual Review results in posts with notable errors getting fixed.In practice, people disagree about what counts as an error, or what errors are particularly egregious. Some errors are quick for an author to fix, some are more gnarly and maybe the author disagrees about the extent to which they are an error.The solution we’ve found for now is to make reviews more prominent on Best Of LessWrong posts, and try to aim for a world where if there is major disagreement, controversy or important considerations about a post, future people will see that disagreement.Currently we do this by including a one-line comment wherever the Spotlight showsup. We may invest more in that over time. This also means if you wrote a review that got 10+ karma, it’s probably worth optimizing the first line to convey whatever information you’d like someone casually skimming the site to read. You can look over the current leaderboard for reviewers to get a sense of which of your reviews might be worth polishing.[1]If you know of someone who’s already written a blogpost or other public response to some works, it’d be helpful to write a short review linking to it and explaining it’s most significant takeaways.The Top 50 as of Jan 1We don’t continuously update the results of the nomination votes, to discourage strategic voting. But, here were the results as-of a couple weeks ago.You might want to note both whether there are posts you think are over/underrated that you want to write in support of. Posts need at least 1 review to make it to the final voting phase.#TitleReviewers1Alignment Faking in Large Language Modelsryan_greenblattJan_KulveitjohnswentworthMarcelo Tibau2On greenJoe CarlsmithRaymond Douglas3Believing InAnnaSalamonBen Pace4The hostile telepaths problemValentineValentineMartin RandallRubyGordon Seidoh WorleyHugo LLucie Philippon5Reliable Sources: The Story of David GerardTracingWoodgrainsThomas KwaScrewtape6Overview of strong human intelligence amplification methodsTsviBTkave7The case for ensuring that powerful AIs are controlledryan_greenblattAlexa Pan8The impossible problem of due processmingyuanAnnaSalamon9Deep atheism and AI riskJoe CarlsmithNO REVIEWS[2]10NeutralitysarahconstantinAnnaSalamon11And All the Shoggoths Merely PlayersZack_M_DavisMartin RandallSeth Herd12Truthseeking is the ground in which other principles growElizabethSherrinford13My Clients, The LiarsymeskhoutScrewtape14Gentleness and the artificial OtherJoe CarlsmithNO REVIEWS[2]15″How could I have thought that faster?”mesaoptimizerNO REVIEWS[2]16Safety isn’t safety without a social model (or: dispelling the myth of per se technical safety)Andrew_CritchAlexa PanZac Hatfield-Doddsthe gears to ascension17What Goes Without SayingsarahconstantinZac Hatfield-Dodds18Repeal the Jones Act of 1920ZviThomas Kwa19Would catching your AIs trying to escape convince AI developers to slow down or undeploy?BuckAlexa Pan20On attunementJoe CarlsmithZac Hatfield-Dodds21There is way too much serendipityMalmesburyGordon Seidoh Worley22The Summoned Heroine’s Prediction Markets Keep Providing Financial Services To The Demon King!abstractapplicScrewtapeabstractapplic23Towards a Broader Conception of Adverse SelectionRicki HeickleneggsyntaxNathan Young24The Field of AI Alignment: A Postmortem, and What To Do About ItjohnswentworthThomas Kwa25Arithmetic is an underrated world-modeling technologydynomightBen Pace26“Alignment Faking” frame is somewhat fakeJan_KulveitJan_Kulveit27Simple versus Short: Higher-order degeneracy and error-correctionDaniel MurfetDaniel MurfetZack_M_Davisniplav28’Empiricism!’ as Anti-EpistemologyEliezer YudkowskyBen Pace29Catching AIs red-handedryan_greenblattBuck30A Three-Layer Model of LLM PsychologyJan_KulveitGunnar_ZarnckeJan_Kulveit31My AI Model Delta Compared To ChristianojohnswentworthRuby32Superbabies: Putting The Pieces TogethersarahconstantinNO REVIEWS[2]33AI catastrophes and rogue deploymentsBuckBuck34Circular Reasoningabramdemskiplex35ThresholdingReview BotScrewtape36Interpreting Quantum Mechanics in Infra-Bayesian PhysicalismYegregVanessa Kosoy37Preventing model exfiltration with upload limitsryan_greenblattNoosphere8938shortest goddamn bayes guide everlemonhopeScrewtape39Transformers Represent Belief State Geometry in their Residual StreamAdam ShaiAdam Shai40Why I’m not a BayesianRichard_NgoRichard_NgoRichard Korzekwa Nathan Young41Hierarchical Agency: A Missing Piece in AI AlignmentJan_KulveitVanessa Kosoy42Being nicer than ClippyJoe CarlsmithNO REVIEWS[2]43You don’t know how bad most things are nor precisely how they’re bad.Solenoid_EntitySolenoid_Entity44Struggling like a ShadowmothRaemonRaemon45Why Don’t We Just… Shoggoth+Face+Paraphraser?Daniel KokotajloDaniel Kokotajlo46The Inner Ring by C. S. LewisSaul MunnNathan Young47Anvil ShortageScrewtapeScrewtapeThomas KwaLorxus48Raising children on the eve of AIjuliawiseNO REVIEWS[2]49Priors and PrejudiceMathiasKBScrewtape50[Intuitive self-models] 6. Awakening / Enlightenment / PNSESteven Byrneslsusr ^(Note, these karma scores subtract your own self-upvote)^Posts without reviews won’t appear in Final Voting PhaseDiscuss ​Read More

Related Posts

Leave a Reply

Your email address will not be published. Required fields are marked *