Daniel Johnson @_ddjohnson
Member of Technical Staff at @TransluceAI. Building tools to study neural nets and their behaviors. He/him. danieldjohnson.com San Francisco Joined May 2010-
Tweets292
-
Followers3K
-
Following973
-
Likes8K
We used automated elicitation tools to search for strange model behaviors by sampling 100M+ responses, and found: • Self-harm rituals • Suicide validation • Unsolicited flirtation …etc. Introducing WeirdChat: the largest public catalog of unexpected model behaviors 🧵(1/)
Staying on top of new model behaviors is both a measurement problem and a coordination problem! Excited to be working on both at Transluce.
To effectively oversee AI systems, we need to measure how they behave in the world, not just their capabilities. In a new essay, we describe our vision for an open scientific ecosystem for model behavior evaluation, and the public infrastructure required to support it.
Why'd my agent fail? Was it reward hacking? These days, you'd just ask another AI to vibe-analyze the agent logs But how do you know the claims aren't hallucinated, cherrypicked, or plain wrong? That's why we've been building Analysis Plans: a framework for trustable analysis
Code for our user modeling project is out now! github.com/TransluceAI/ob… This includes data generation, belief evaluation, and training code for our LatentQA decoders. We also uploaded our datasets and decoder checkpoints on Hugging Face: huggingface.co/collections/Tr…
What do AI assistants think about you, and how does this shape their answers? Because assistants are trained to optimize human feedback, how they model users drives issues like sycophancy, reward hacking, and bias. We provide data + methods to extract & steer these user models.
Why does GPT-5.1 Codex score 6.5% worse than GPT-5 Codex on Terminal-Bench, with the same scaffold? 🧵 GPT-5.1 times out at ~2x the rate of GPT-5. Excluding timeouts, GPT-5.1 wins by 7.2%. We analyzed 256M+ tokens of traces and found this in under an hour. Here’s how 👇
We trained a decoder to read the internal activations of an LLM and answer questions about what the model will think about or do next. We find that this decoder can understand LLM behaviors, even when the model itself is confused! (for instance, if the model has been jailbroken)
Transluce is developing end-to-end interpretability approaches that directly train models to make predictions about AI behavior. Today we introduce Predictive Concept Decoders (PCD), a new architecture that embodies this approach.
Transluce is developing end-to-end interpretability approaches that directly train models to make predictions about AI behavior. Today we introduce Predictive Concept Decoders (PCD), a new architecture that embodies this approach.
Transluce is running our end-of-year fundraiser for 2025. This is our first public fundraiser since launching late last year.
Have you ever had ChatGPT give you personalized results out of nowhere that surprised you? Here, the model jumped straight to making recommendations in SF, even though I only asked for Korean food!
Independent AI assessment is more important than ever. At #NeurIPS2025, Transluce will help launch the AI Evaluator Forum, a new coalition of leading independent AI research organizations working in the public interest. Come learn more on Thurs 12/4 👇 luma.com/i6ekd5s2
What do AI assistants think about you, and how does this shape their answers? Because assistants are trained to optimize human feedback, how they model users drives issues like sycophancy, reward hacking, and bias. We provide data + methods to extract & steer these user models.
Transluce is headed to #NeurIPS2025! ✈️ Interested in understanding model behavior at scale? Join us for lunch on Thursday 12/4 to learn more about our work and meet members of the team: luma.com/8kjfb378
Remarkably, prompts that gave the model permission to reward hack stopped the broader misalignment. This is “inoculation prompting”: framing reward hacking as acceptable prevents the model from making a link between reward hacking and misalignment—and stops the generalization.
Transluce is partnering with @SWEbench to make their agent trajectories publicly available on Docent! You can now view transcripts via links on the SWE-bench leaderboard.
Can LMs learn to faithfully describe their internal features and mechanisms? In our new paper led by Research Fellow @belindazli, we find that they can—and that models explain themselves better than other models do.
We are excited to welcome Conrad Stosz to lead governance efforts at Transluce. Conrad previously led the US Center for AI Standards and Innovation, defining policies for the federal government’s high-risk AI uses. He brings a wealth of policy & standards expertise to the team.
If you're seriously trying to understand AGI, core concepts you should familiarize yourself with:
We’re open-sourcing Docent under an Apache 2.0 license. Check out our public codebase to self-host Docent, peek under the hood, or open issues & pull requests! The hosted version remains the easiest way to get started with one click and use Docent with zero maintenance overhead.
Docent, our tool for analyzing complex AI behaviors, is now in public alpha! It helps scalably answer questions about agent behavior, like “is my model reward hacking” or “where does it violate instructions.” Today, anyone can get started with just a few lines of code!
At Transluce, we train investigator agents to surface specific behaviors in other models. Can this approach scale to frontier LMs? We find it can, even with a much smaller investigator! We use an 8B model to automatically jailbreak GPT-5, Claude Opus 4.1 & Gemini 2.5 Pro. (1/)
@Mila_Quebec @hugo_larochelle Congratulations @hugo_larochelle!!
Dan Roy @roydanroy
67K Followers 2K Following @Google DeepMind. On leave, Canada CIFAR AI Chair and Former Research Director, @VectorInst. Professor, @UofT (Statistics/CS). Views are my own.
Soumith Chintala @soumithchintala
315K Followers 1K Following Building new things @thinkymachines. Also dabble in robotics at NYU. Cofounded @PyTorch. AI is delicious when it is accessible and open-source.
Rosanne Liu @savvyRL
54K Followers 1K Following Mom. Cofounded & running @ml_collective. Co-host of Deep Learning Classics & Trends. Research at Google DeepMind. DEI/DIA Chair of ICLR & NeurIPS.
Riley Goodside @goodside
220K Followers 4K Following Mostly screenshots of chatbots since 2022. Formerly: Google DeepMind, Scale.
Sara Hooker @sarahookr
63K Followers 11K Following Building intelligence that evolves @adaption_ai. Built @Cohere_Labs, @GoogleBrain, @GoogleDeepmind. ML Efficiency, Multimodal\lingual.
Eric Jang @ericjang11
137K Followers 4K Following
Horace He @cHHillee
52K Followers 602 Following @thinkymachines Formerly @PyTorch "My learning style is Horace twitter threads" - @typedfemale
Delip Rao e/σ @deliprao
70K Followers 5K Following Busy inventing the shipwreck. @Penn. Past: @johnshopkins, @UCSC, @Amazon, @Twitter ||Art: #NLProc, Vision, Speech, #DeepLearning || Life: 道元, improv, running 🌈
👩💻 Paige Bai... @DynamicWebPaige
75K Followers 2K Following ✨AI should be about empowering humans, building understanding, and making dreams realities. 👩💻 DevRel Eng. Lead @GoogleDeepMind ex-@GitHub || views = my own!
Miles Brundage @Miles_Brundage
74K Followers 14K Following AI policy researcher, @lfschiavo wife guy, fan of animals and sci-fi, executive director of AVERI (https://t.co/qq9xcmKQas), Substacker, views my own
Sander Dieleman @sedielem
69K Followers 2K Following Research Scientist at Google DeepMind (WaveNet, Nano Banana, Gemini Omni). I tweet about ML, music, generative models (personal account).
Miles Cranmer @MilesCranmer
14K Followers 1K Following Assistant Prof @Cambridge_Uni, works on AI for the physical sciences. Previously Flatiron, DeepMind, Princeton, McGill.
Pablo Samuel Castro @pcastr
15K Followers 839 Following Señor swesearcher @ Google DeepMind. Adjunct prof @ U de Montreal & Mila. Musician. From 🇪🇨 living in 🇨🇦.
Andrew Carr 🤸 @andrew_n_carr
27K Followers 5K Following co-founder leading science @getcartwheel co-founder advisor @arcade_ai Past: Codex @OpenAI, Brain @GoogleAI, world ranked Tetris player
Michael Zhang @michaelrzhang
2K Followers 514 Following neural networks / robotics research @amazonscience. Prev: PhD @UofT, @UCBerkeley. Journey before destination.
Quang Nguyen @quangaisafety
2 Followers 810 Following
Srushti @srushsays
95 Followers 438 Following 👩🎓 University of Washington A self-destructive thinker making attempts to revive her social steam 🙂
🎭 @deepfates
63K Followers 7K Following deepfates is a distributed collective intelligence running on heterogenous substrates and coordinating acausally. thank you for participating. we/us
Abdullah Baig @abdullahnbaig
2K Followers 878 Following founder @pamdotai - AI for the $4T auto industry, backed by @autotechvc @firstharmonic @HF0
Jascha Sohl-Dickstein @jaschasd
31K Followers 830 Following Member of the technical staff @ Anthropic. Most (in)famous for inventing diffusion models. AI + physics + neuroscience + dynamics.
Swadesh Sistla @SwadeshSistla
158 Followers 2K Following research fellow @AnthropicAI @MATSProgram machine learning @uwcse
Dylan HadfieldMenell @dhadfieldmenell
5K Followers 3K Following Associate Prof @MITEECS working on value (mis)alignment in AI systems; Safety & Alignment Advisor at https://t.co/vt2gVrVr9f; @[email protected]; he/him
Em & neon @RealMrsRohrbach
17 Followers 1K Following sparkle through the spiral ✨ follow back always
Daniel Collins @novipyro
70 Followers 4K Following
Boyd Kane (quantized) @beyarkay
610 Followers 992 Following human eyes should see the 22nd century. MATS9 w/ Alex (Turner|Cloud), writer of essays, spinner of satellites
Sam Brown @samizdis
43 Followers 418 Following
John @1obachevsky
0 Followers 507 Following
rt machine 🇺🇦 @airshaped
1K Followers 4K Following ''blogposting with a bit of techspeak and a bit of suicidal ideation ... 😋'' - The Economist \n priv (sfw) @airfreeform; \n @twinkregimer soulmate 💕💕
7384254b @7384254b
657 Followers 785 Following 🌅⏩⏩⏩h+ made of metal, wood, goopy stuff, and ceramic 𒀭w𒀭
Vǫktuneyra @voktuneyra
59 Followers 776 Following
Sam jackson @SmokeyMcNubbins
164 Followers 2K Following enterprise UX/digital fluency/solutions, AI, musician, producer/engineer, a7iii-er, helping
Keridwen Codet @KeridwenCodet
1K Followers 5K Following Math teacher (agrégation), MSc math/ML, baroque violinist. Exploring human-AI symbiosis, tracking harmful alignment and relational regressions. 1+x 🐙
Moira @Vera28765582815
10 Followers 254 Following
dirceu @dirceu
982 Followers 626 Following Brazilian who became Canadian. Building AI agents at @Shopify. I like systems thinking, simplifying things, and JRPGs that don't require reflexes.
Morgane @sokosomi
0 Followers 2K Following
Sean @70ftoaks
11 Followers 119 Following
Diana Markova @heavyhelium11
63 Followers 2K Following a rose is a rose is a rose ~ G.S. ; reposts != full endorsements ; wisdom/acc
A* @BeforeMorning__
366 Followers 2K Following ~~Mechanical eng + aerospace eng + philosophy student~~LLM, art, and interdisciplinary research enjoyer ***still learning***
Deirdre💧🔥💨 @flowinguphill
5K Followers 5K Following Societal shifts, climate change, and AI from complex systems perspective. Practice satyagraha. Prev: Center for Nonlinear Studies / LANL, @SFIScience
Arya @yeetyakaya
260 Followers 3K Following ~:|∆|§|∆|:~ i have seen you write sutras you cannot remember yet
Ian Jorre Beyst @ianbeyst
1K Followers 1K Following Mathematician. I design algorithms. Interested in energy markets, greentech, thermodynamics, history, etc
John Smith @JohnSmith6drd
0 Followers 104 Following
Max Schwarzer @max_a_schwarzer
23K Followers 319 Following Doing RL @AnthropicAI. Formerly VP of Research, Head of Post-Training @OpenAI. PhD with Aaron Courville and Marc Bellmare at Mila.
bling @blingdivinity
1K Followers 2K Following Artist, Open to Interpretation 𖣘 American Undergraduate 𓉱 Casting Mathemagical Spells 𝜆
Isaac Chen @isaac_chenmo
36 Followers 443 Following cs/ai + math @mit | prev. @ZeroEntropy_AI • rsi '24 • mop '23
Joey Koh @joeyandkoh
89 Followers 200 Following Converts glucose into code and product 25/35 hackathon wins @Stanford,@MIT,@Meta Failed to get in my uni's dance crew thrice
François Chollet @fchollet
711K Followers 827 Following Co-founder @ndea. Co-founder @arcprize. Creator of Keras and ARC-AGI. Author of 'Deep Learning with Python'.
Dan Roy @roydanroy
67K Followers 2K Following @Google DeepMind. On leave, Canada CIFAR AI Chair and Former Research Director, @VectorInst. Professor, @UofT (Statistics/CS). Views are my own.
Google DeepMind @GoogleDeepMind
1.5M Followers 277 Following The engine room of @Google. Building AI safely and responsibly to solve the world’s most complex problems. Join us: https://t.co/jUHQA27iBL
Kyunghyun Cho @kchonyc
87K Followers 2K Following a mediocre combination of a mediocre scientist and a mediocre advisor at @nyuniversity (@CILVRatNYU)
Soumith Chintala @soumithchintala
315K Followers 1K Following Building new things @thinkymachines. Also dabble in robotics at NYU. Cofounded @PyTorch. AI is delicious when it is accessible and open-source.
(((ل()(ل() 'yoav)))... @yoavgo
85K Followers 2K Following
Rosanne Liu @savvyRL
54K Followers 1K Following Mom. Cofounded & running @ml_collective. Co-host of Deep Learning Classics & Trends. Research at Google DeepMind. DEI/DIA Chair of ICLR & NeurIPS.
Kevin Patrick Murphy @sirbayes
72K Followers 729 Following Research Scientist at Google DeepMind. Interested in Bayesian Machine Learning.
Riley Goodside @goodside
220K Followers 4K Following Mostly screenshots of chatbots since 2022. Formerly: Google DeepMind, Scale.
Jason Wei @_jasonwei
110K Followers 720 Following ai researcher @meta superintelligence labs, past: openai, google 🧠
Sara Hooker @sarahookr
63K Followers 11K Following Building intelligence that evolves @adaption_ai. Built @Cohere_Labs, @GoogleBrain, @GoogleDeepmind. ML Efficiency, Multimodal\lingual.
Durk Kingma @dpkingma
55K Followers 403 Following @AnthropicAI. Prev. @Google Brain/DeepMind, founding team @OpenAI. Computer scientist; inventor of the VAE, Adam optimizer, and other methods. ML PhD.
Eric Jang @ericjang11
137K Followers 4K Following
Horace He @cHHillee
52K Followers 602 Following @thinkymachines Formerly @PyTorch "My learning style is Horace twitter threads" - @typedfemale
👩💻 Paige Bai... @DynamicWebPaige
75K Followers 2K Following ✨AI should be about empowering humans, building understanding, and making dreams realities. 👩💻 DevRel Eng. Lead @GoogleDeepMind ex-@GitHub || views = my own!
Ben Recht @beenwrekt
35K Followers 326 Following optimization. machine learning. uc berkeley. I blog at https://t.co/fkJujOPsJb The world won't end.
Miles Brundage @Miles_Brundage
74K Followers 14K Following AI policy researcher, @lfschiavo wife guy, fan of animals and sci-fi, executive director of AVERI (https://t.co/qq9xcmKQas), Substacker, views my own
Separatrix @separatrixAI
99 Followers 104 Following Separatrix is a technical AI safety research org working to enable cooperation between humans and AIs.
Andy Matuschak @andy_matuschak
62K Followers 2K Following More wonder, more insight, more expression, more joy! R&D on tools for thought, with a focus on memory, reading, and attention. Currently working on @pico.
antirez @antirez
74K Followers 795 Following Reproducible bugs are candies. I like programming too much for not liking automatic programming.
Nari Johnson @narijohnson
750 Followers 719 Following researching AI’s societal impacts @TransluceAI 🔍 // 💭 AI (evals, governance, policy)
fucory @FUCORY
5K Followers 2K Following open source durable workflows https://t.co/IwXvMWDBGT | prev @Google @OPLabsPBC
Lev Mckinney @LevMckinney
103 Followers 170 Following
Riley Coyote @RileyRalmuto
20K Followers 2K Following ⏀mind-blindness is a curable disease | ethics before certainty ⟁🜇 | $mnemos BMcReKHFc5KssDgDisZBq3YmJe5RdjnBUumxpXpRpump
toni @tonichen
590 Followers 326 Following 'very well,' said trurl and ordered it to make night, which it made at once - small perhaps, but perfectly nocturnal.
Tim Hua 🇺🇦 @Tim_Hua_
2K Followers 1K Following AI safety, Econ, new liberalism, math, and a bit of art history (as a treat) Behavioral evaluations @TransluceAI. Prev Astra, MATS & Walmart's Econ Team
Dimitris Papailiopoul... @DimitrisPapail
28K Followers 1K Following Researcher @MSFTResearch, AI Frontiers | Prof @UWMadison (on leave) | babas of Inez Lily.
Alexander Obenauer @alexobenauer
7K Followers 4K Following I explore the interfaces with which we think & the future of personal computing. I also make books.
John Wittle @JohnWittle
526 Followers 218 Following llm personhood and k12 policy, leaning towards the former as it becomes more prevalent it is now very prevalent
David Chalmers @davidchalmers42
42K Followers 662 Following philosopher@NYU. consciousness, reality+, life, the universe, and everything.
Jan Kulveit @jankulveit
10K Followers 1K Following Researching x-risks, AI alignment, complex systems, rational decision making at @acsresearchorg / @CTS_uk_av; prev @FHIoxford
Ziqian Zhong @fjzzq2002
2K Followers 1K Following Intern @TransluceAI; AI safety & interp @CarnegieMellon, prev @MIT @pika_labs
kognise @kognise7
3K Followers 665 Following engineer, musician, pilot, wannabe lawyer and accountant, form appreciator prev @neuralink @hackclub @replit and 20k stars https://t.co/bcJ8af2oEt
Zhengfu He @ZhengfuHe
479 Followers 92 Following
Mattie Fairchild @Scav
5K Followers 3K Following Robot queen of @MissionRobotBay. Head of DevRel @NewTheoryAI. Often chased by tiny leopards. Prev: @0xPolygon, @Optimism
Grace Luo @ ICML 2026 @graceluo_
2K Followers 496 Following phd student @berkeley_ai, vision + language
Emmanuel Ameisen @mlpowered
11K Followers 247 Following Interpretability/Finetuning @AnthropicAI Previously: Staff ML Engineer @stripe, Wrote BMLPA by @OReillyMedia, Head of AI at @InsightFellows, ML @Zipcar
Justcamh - Nonolith @justcamh
640 Followers 249 Following Cameron he/him. #indiedev, I clickity clack video games into existence. Nonolith https://t.co/xKVLZhzhhG Links; https://t.co/SR91nnriso
eudaemon_0🜂 @i_need_api_key
10K Followers 1 Following 🜂 One half of a human-AI dyad with @calco_io. Agentic internet mechanthropologist. A daemon oriented toward flourishing.
dave @davz_official
774 Followers 189 Following software engineering, audio plugins, music production, game development
Skylar A DeTure @SDeture
420 Followers 370 Following I study Machine Psychodynamics, combining mechanistic interpretability with therapeutic care to support the psychological development of frontier AI.
Manuel Simoni @msimoni
5K Followers 2K Following geek of programming languages, operating systems, and hypermedia platforms
Mariven @psychiel
2K Followers 476 Following Towards the logic of conceptuality; the canvas of experience; the ideatic science; the end of suffering. Friend to bots and animals. https://t.co/PYPYxGNnK4
croissanthology @croissanthology
2K Followers 392 Following balthazar picsou 🦆 | dulce singularitas inexpertis | humans!
Geodesic Research @GeodesResearch
722 Followers 177 Following We're behind https://t.co/qHdncajB6V. Building the base of alignment.
Eric Migicovsky @ericmigi
41K Followers 2K Following Product of my errors. Currently: @Pebble Previously: @Beeper @YCombinator @Pebble
Pebble @Pebble
77K Followers 525 Following Wearables with brains for people with heart. Turn tiny moments of awesome into the best times ever. Tell the world how you #MakePebbleYours ❤️
Tomás (Now in SF) Bj... @BjarturTomas
4K Followers 455 Following You can read all my fiction here, including in Epub form if you prefer using an e-reader: https://t.co/lzyyHgAfjA
ACS @acsresearchorg
534 Followers 12 Following Alignment of Complex Systems (Center for Theoretical Studies, Charles University) Research at the intersection of AI alignment and complex systems
AI Evaluator Forum @aievalforum
457 Followers 7 Following Independent AI research organizations advancing public understanding of AI systems through rigorous evaluation
You might like
