San Francisco, CA
Joined July 2022
two years ago all my friends were bi-curious and now they’re all bio-curious
9
103
6,302
makes so much sense that the reason why Liz refuses to admit to lying in any capacity is because it would destroy her narcissistic self image of being someone truthful / “someone who always says what they believe” she and billy are both psychologically unable to admit any evidence that contradicts their self delusions, which is an insane self protection strategy and the only way two people can remain this far removed from reality
You Can See Everything: a Psychoanalytic Interpretation One of Nathan Fielder’s gifts is the ability to curate a conversational space so inviting that a speaker reveals more of themselves than they initially intended. In You Can See Everything (Oppenheim, 2026), Elizabeth Holmes and her husband, Billy Evans, experience this firsthand, and the stillness lasts long enough that you can observe their personality structures attempt to retain their shape on camera, with moments of breaking frame. Holmes smiles with wide eyes and smile for extended (one might say uncanny) stretches while Billy’s eyes stay intense. Theirs is a story about lying, intense love, and a woman who insists she has never done the one thing the world convicted her of: lying at a massive scale. I watched the film as a clinician first and a moviegoer second (okay, and a serious Nathan Fielder fan a CLOSE third). What follows is my read through Nancy McWilliams’s personality theory and through Fonagy’s work on mentalization. It is interpretation, not courtroom psychology, and it rests on Holmes’s own words as reported from the theatrical cut. Holmes does not stop at simply disputing a charge. She denies the category of the act itself. “I’ve never tried to deceive anyone,” she tells Fielder. Asked whether she has ever lied in her adult life, replies: “I don’t think so.” Fielder, for once stripped of the blank slate questioning style, asks: “Liz, what?” Billy, almost fondly: “She’s weird.” Liz is famous for that thing she’s allegedly “never done”: lying. McWilliams (2011) places shame at the center of narcissistic organization. Grandiosity, in her account, is not swagger for its own sake. It is a structure built to keep annihilating shame out of awareness. After all: admit a lie and you admit a crack in the self-image that was supposed to make you invulnerable. Rewrite history instead, and the self stays intact for one more day. Holmes’s insistence that she has always “tried to say what I believe” fits that defense with almost textbook neatness. Belief becomes the solvent in which other people’s experience of deception is supposed to dissolve. Film is a great medium for this interview, as we can see Holmes’s expression remain warm throughout the interview: the smile ready, the presentation of approachability intact even during serious, upcoming incarceration. Nancy McWilliams describes how narcissistic personalities often rely on charm and a carefully tended surface to pull admiration, the supply that keeps the grandiose self afloat. The warm eyes function as a mirror held up to the room: "see how sincere I am, and reflect that back. I'm here as a positive force." Fonagy and Target (1997; see also Fonagy, Gergely, Jurist, & Target, 2002) describe mentalization as the capacity to hold another person’s mind in mind: to recognize that other people have beliefs, feelings, and intentions that may not match your own, and that those mental states are what organize their behavior. Theory of mind is the everyday language for the same skill. Without it, other people collapse into props in your story. Holmes keeps treating other minds that way, insisting she has never tried to deceive anyone, as if her own sincerity cancels what investors, patients, and jurors experienced. She misses Fielder’s joke about hosting a show (“Nathan for You,” which I recommend you watch) on business scams while Liz is actually convicted of an actual scam. Holmes remarks that she had not thought about that coincidence until he said it. Later, Fielder explains that he perceives Liz to use language in a technically honest way, regarding her intent, but in a manner which would obscure and confuse others. To illustrate this, he loads up and plays the seriously old school “Who’s on First” Abbott and Costello comedy skit, a famous exercise in linguistic miscommunication. She says, plainly, “I didn’t pick it up.” Humor requires a brief exit from your own perspective. You must see how the other person is experiencing a situation, from their vantage point, which at least in this instance, she’s unable to do. Billy’s interruptions of Christian Holmes (Elizabeth’s father) show an unwillingness to entertain other perspectives, rather than a failure to perceive them. When Fielder asks about Sunny Balwani, (Liz’s business partner in her Theranos days) Billy jumps in. When the father names a turning point (”That’s an easy one,” her sexual assault at nineteen) and identifies Balwani as the mastermind who manipulated her, Billy derails and enters the conversation again, never allowing Nathan and Liz’s father to examine the traumatic background. The conversation stops being about what Christian thinks. It becomes about what can be admitted into the shared mythos that Billy is building. McWilliams already assumes that narcissistic structure forecloses true empathy; add impaired mentalization and you get someone who can “invent from a cage,” as Liz says she will do from prison, who can “build monuments of love” (as both Liz and Billy say that Liz will do for Billy), and fail to register that the people outside the shared culture of the couple do not wish to play along inside the “bubble.” Prison might be expected to puncture the bubble of grandiosity, but for Holmes it becomes merely a new stage. “From wherever I am,” she says, “I’m going to continue to work inventing, even if it’s inventing from a cage.” McWilliams writes that grandiosity and depletion are two faces of the same personality organization: the depleted self feels empty and small. The grandiose self swells to cover that emptiness. When the covering story is threatened, the person often doubles down rather than mourning. Holmes still speaks with sweeping ambition and almost no qualifications, so the cage doesn’t revise the script, it’s just a new stage design. I find the missing sense of humor clinically useful here. A joke asks you to tolerate some shame, a little being-seen-as-ridiculous. Narcissistic defenses treat that request as an attack. So the room goes flat. Fielder offers play. She offers sincerity. Those are different currencies, and they do not convert. Christian Holmes’s interview is one of the film’s most revealing sequences precisely because Billy will not allow it to unfold, uninterrupted. He meddles and sabotages it at every turn. McWilliams notes that narcissistic defenses often crystallize around unbearable injury, with a grandiose self constructed over shame too great to feel. The “Invincible inventor” identity looks like a fortress raised after negative experiences. Sunny was once Elizabeth’s right hand man, with Billy now fulfilling that role, deciding who gets to access to Liz. Caution is necessary here, because the presence of a trauma, such as an assault, does not excuse fraud, and it does not make Billy (or Sunny) into cartoonish villains. Billy reports that Liz claims “every action she takes for as long as she lives will be building monuments of love” to Billy. In Fielder’s reconstruction of a prison visit (Amanda Seyfried standing in for Holmes, staged from Fielder’s memory of the conversation), she describes the two of them building monuments of love for each other and says, “I have conversations with him in my head.” Billy, from the outside of prison says: “I know what she’d want.” Relatedly, McWilliams describes the narcissistic cycle of idealization, in which the other person is prized as a reflecting surface more than as a separate subject. Here the reflecting works in both directions. Each carries a working model of the other so complete that speech becomes optional. They have built a digital machine of each other, enmeshed in idealisms, rehearsing what the beloved would say until the beloved becomes a program running in the head. That arrangement can feel like profound intimacy from the inside. From outside, it looks like two monuments facing each other across visiting glass, neither quite able to see a living person who might surprise them. McWilliams distinguishes narcissistic from psychopathic (or antisocial) personality organizations with unusual clarity. The narcissist is organized around shame and the hunger to be admired. The psychopath is organized around omnipotent control and often takes satisfaction from putting one over on others. Visibility is optional for the second type, while results are not. The narcissist is a convenient monkey’s paw for the psychopath, because the narcissist must be SEEN winning to be fulfilled, while the psychopath only needs the win itself. I should double down in my disclaimer that this is simply for entertainment purposes, not medical therapeutic diagnosis; simply commentary one a nearly 3-hour film's character depictions. There’s a fascinating subplot in which Fielder tries to identify whether Liz is aware that Billy Evans asked for substantial monetary compensation as a condition of his and Liz’s participation in the documentary (which is extremely unorthodox, as far as documentaries are concerned). Nathan verbally recounts the timing of Billy’s request, and notices that Liz leaves the room right before Billy makes his proposition. Does this indicate she’s aware of what’s coming, and she wishes to maintain plausible deniability? Her response when Nathan asks her at prison indicates that perhaps she was aware, as she maintains she allows Billy to handle financial matters. This is not a shocked reaction, but one of deflection. By the end of the film (SPOILER) we learn that Billy has relinquished his request for royalties, indicating at the very least, he and/or Liz thought better of the inevitable optics nightmare. However, I think it’s a bit too late, as it’s almost the worst of both worlds to have asked for the money, revealing the willingness to override typical documentary policy, while not even receiving the literal compensation in exchange for the bad press. Liz and Billy both walk away looking worse for wear, in this documentary. The form a culture of two; both frenetic, entrepreneurial forces, with Billy willing to embody the independent, disagreeable businessman persona and Liz appearing as the “good cop.” It’s difficult to know just how much this pattern may have represented the dynamic between Liz and Sunny Balwani at Theranos. It’s a coupling of masculine/feminine approaches to winning at business, like a marriage of “Lean In” and “The Art of the Deal.” The audience can see the 1-2 punch dynamics at play. You Can See Everything is a strange title for a story about people who cannot see each other. Fielder sees a great deal. Holmes sees a continuous invention project. Billy sees a monument under construction. Christian sees a daughter damaged by assault and by Sunny Balwani. The audience sees a woman who will not grant that she has ever lied, and a marriage that has turned mutual idealization into a closed loop, a “culture of two.” Mentalization would have interrupted the loop with new data. So would a joke she could laugh at, or a sentence that begins, I deceived people, and here is how. There is no sense of authenticity, just a veneer of sincerity performed at cinematic scale, warm eyes aimed at the lens, and the promise that invention continues even from a cage. The clinical moral of the tale is that if someone cannot hold your mind as separate from theirs, your experience will be missed, or at the very least, misrepresented. What does a friend or close connection do, when presented with evidence that hurts? How do they respond when they’ve hurt you? Is there an apology, or this there an alternate account of reality?
1
1
22
1,546
Give me a trillion dollars and I'll solve all disease in 5-10 years.
Largely agree with these thoughts, and have some perhaps discordant points to add. - Solving biological diseases in the next few years will benefit mostly from the usefulness of the language models, substantially more than molecular ML. - Proponents of the bitter lesson for biological models (not LLM) aren't wrong, just (possibly pathologically) too early. We don't have enough data, and the data we can collect quickly is probably not the right kind (we may get lucky with biology being easy, but it's usually not). - I think just scaling models and public (passively accumulating) data is a cost ineffective direction (it's like scaling dense NNs on the internet). Today's model outputs are as good as GPT2. A single bioinformatician with linear regression, sometimes even without any additional data, can beat these models in many applications (people rarely benchmark AI advances against a simple Potts model). The most promising approaches for molecular models are innovating heavily on architecture (e.g. Baker Lab, Alphafold, etc). This may automate/speed up with LLMs. - Few companies (hopefully Dyno included) are betting that we've triangulated enough pieces from multiple experimental channels to predict (with scale) how things translate in humans. Having worked on translating in vitro things to primates, I'm very cautious with the idea that current models, trained on current public data, can do anything more than manipulating already working solutions, either explicitly or through implicit remixing of what has worked. It may still work, but the underlying bets offer no specific reason why there is any epistemic risk advantage other than being "made with AI". - Short term, there is more promise in combining molecular models with LLMs rather than scaling them. The two sources of lift in AI discovery elsewhere has been (1) Breaking Silos: AI synthesizes a bunch of knowledge/expertise that is rarely present in the same room (2) Data mining : AI pushes through volumes of data, with an intelligent eye, that humans or simple algorithms simply can't scale or adapt to. I'm curious what will come out of multimodal training of LLMs and biological data, which seems to be something the big players are pursuing. I think latent reasoning in the language (or better yet, + code) domain will reveal a lot more biology than scaling on nucleotides or proteins alone. - My view is that the next opportunity of combining lab and AI will be massive-scale hypothesis testing with LLMs. Currently a multiplexed assay generally tries to test a handful of ideas because the humans are bandwidth limited in designing properly powered experiments. LLM's can scale this very well. I have no doubt lots of good news will come on the impact of AI in biology in the next five years. I can't tell what paradigm will win, but my prediction is that it will not be the same paradigm imported from the LLM land (which will get a ton of funding anyways, because it's easy to pattern match for tech VCs).
3
8
1
158
7,505
Melissa Du retweeted
I’ve left Anthropic after more than a year of working on Claude Code and seeing it scale larger than I ever imagined. I have really loved working at Ant and believe deeply in the mission, and most of all, the idea of a race to the top. The idea that if we work hard and responsibly, we can raise the bar and inspire others to great outcomes. And if we all push each other to be better, things can be radically good. I will carry the spirit of this with me forever. My colleagues at Anthropic are the most talented, thoughtful, competent, and earnest people I have ever met. I am glad they are at the wheel and trust them deeply. To the community on X, thank you for everything. You’ve made this the most fun (and sometimes stressful) job I’ve ever had. For me, it is time for a bit of a break before diving into something new. If you want more excellent Claude Code content, you should follow @addyosmani , @lydiahallie, @_catwu, @ClaudeDevs and of course @bcherny who started it all.
2,227
1,109
870
31,441
6,225,831
We’ve developed the world’s most complete megakernel framework. It can perform entire training runs, serve inference, play video games, and even do all three simultaneously in a single kernel.
39
86
79
1,367
188,610
Melissa Du retweeted
Today we are excited to release Artios, a fully open source system for designing functional de novo enzymes. Enzymes present a new frontier for de novo protein design models due to the challenge of coordinating precise chemistry, rather than binding a target. Enzymes are life’s biological machines and by designing them we can truly start to engineer biology. We believe these fundamental capabilities should be built in the open and shared with the community to progress science faster. 🧵 1/6
23
93
11
423
29,715
Omnii-1 is out! We started training a few months ago, and the range of applications we've found that it can tackle has surprised even us. It's been a clear lesson in the power of scale — in parameters, data, context, tasks, modalities, etc. Grateful to have built it with this incredible team.
Introducing Omnii-1, our most powerful genome language model, and a major step toward general biological intelligence In a world first, Omnii-1 designed the first AI-generated bacteriophages to be administered to a human patient with an antibiotic-resistant infection under compassionate use
8
6
58
2,775
I am actually So Impressed by how many new capabilities Omnii-1 was announced with Congrats @RadicalNumerics :)
Introducing Omnii-1, our most powerful genome language model, and a major step toward general biological intelligence In a world first, Omnii-1 designed the first AI-generated bacteriophages to be administered to a human patient with an antibiotic-resistant infection under compassionate use
3
2
66
2,732
I think Nathan Fielder and Elizabeth Holmes are far more similar to each other than the average person or anyone else in that film
14
6
3
433
10,309
Melissa Du retweeted
terry tao: imagine, if you will, a petrifying situation. you have produced a very effective drug, but you do not know how it works. do you trust the drug? now, this is just a hypothetical, but— drug developers:
This is a slide from Terence Tao's 'Math 2.0' public lecture yesterday evening at Caltech. I am highly confident that, if the cure actually worked, almost any Stage 4 cancer patient would tell you that human comprehension of its mechanism is completely irrelevant to them.
47
128
21
1,888
79,820
PIE, the latest Virtual Cell model release, includes language embeddings of curated text describing the perturbation, the cell line, and each gene as part of its input space. Think it’s likely that this approach will be superseded by frontier LLMs trained on perturbation data.
1
2
34
2,540
After thinking about it, I have come to the unfortunate conclusion that frontier labs (and wet lab shops) are all you need.
10
1
157
16,575
Of course, the base model can still write and use tools to abstract away certain functions or capabilities. The tools can still be other predictors or simulators. The reality will still be messy and the most performant systems will likely involve base models with great generalization and tools, both specified and learned.
2
1
1
5
1,533
That said, when base models become better than humans for writing and refining tools, the only thing that matters outside of the frontier lab stack becomes wet lab shops for validation...
1
1
11
926
It's been a while since I've been this excited about a model, it writes better, it writes smarter, it writes shorter, and it's a pleasure to talk to. Bust most important of all, it codes well. It's a good model, sir.
Replying to @claudeai
At its default effort setting, Opus 5.5 delivers frontier results for a fraction of the cost per task, often beating other models running at their highest settings. It also generates output more than 30% faster than Opus 5.
1
1
53
3,044
Melissa Du retweeted
$20k referral fee if we hire someone you send us. Serova is building fully vertical next-gen healthcare, from the lab to the clinic. Backed by Tier 1 investors. Roles open right now: SCIENCE & LAB Bioinformatics Wet lab immunology Wet lab robotics Wet lab high-throughput MANUFACTURING & ROBOTICS mRNA manufacturing Robotics engineering ENGINEERING & PRODUCT ML engineering Software engineering Product engineering Design CLINICAL & PATIENTS Oncologists Nurse navigators Patient relations Clinical operations Translational project managers Serova clinic leads (yes, we're building our own clinics) REGULATORY & FINANCE Regulatory Medical legal Insurance Medical finance OPERATIONS & SPECIAL PROJECTS Email me at [email protected] if you want to chat. Tag someone who should see this. (we had no team photos this is the best I could do)
42
32
6
254
28,671
Melissa Du retweeted
Some news: 5 months ago I joined @coreauto, a small research lab focused on new ways for models to learn. I’m also hiring four interns to do something my younger self would've loved. It started with a pretty unusual project. We set up a handful of small online businesses, and put an agent in charge. We gave each business some initial funding and a seed idea, and off they went, trying to sell real things to real customers while we watched. It’s about as open-ended and unforgiving a task as you can probably give to an agent, with countless decisions to make and effects that take time to become clear. Each business got its own bank account, a debit card to pay suppliers, a web browser, an email inbox, a domain name, cloud infrastructure and a schedule. Every few hours, the agent woke up and decided what to do. It might update a landing page, adjust an ad campaign, reply to a customer, and then go back to sleep. The agents built actual storefronts and apps. They bought ads, attracted visitors, and even made some sales. The agents acted as operators that kept a business moving, but we noticed they never really stepped back to strategize about what might move the needle. If people weren’t buying, why? Was the landing page unappealing? Or maybe the offer itself was the problem? One amusing example was the very first order a customer placed. Our supplier had run out of stock. But it’s a dropshipping business, and thankfully tens of other suppliers carried the same product. Instead of trying another supplier, the agent almost immediately sent the customer a polite apology and attempted to issue a refund. Gah! So close! The agents could handle much of the work of getting a business up and running. What they struggled with was deciding what to change when it wasn’t working. That’s what we want to understand. We also want to see what changes when people take the lead. What does a business operator actually do all day when agents can do so much? What new work does this allow them to do? Which decisions can’t be delegated? — So this winter we're hiring 4 interns as Associate General Managers. Associate Product Manager internships are programs that give early career generalists a product to own. We want to try something similar, but with entire businesses. You’ll create and grow businesses, with agents handling the routine work so you can focus on figuring out what matters. Each AGM gets a budget and our agent infrastructure that’s designed for this. Which ideas to try is up to you. Put more money behind what finds customers and keep refining. Where you step in tells us what agents can't do yet. You'll be in our SF office with me and people I’m so fortunate to work with, like @_arohan_, @joannejang, @MillionInt, @marksaroufim and @juliacvillagra We don't care much where you went to school or what your resume looks like. We’re looking for enterprising people with a drive to create, who’ve made money on the internet before and already use AI extensively to enable their ideas. If that sounds like you: coreauto.com/AGM
102
69
37
1,698
251,043
Melissa Du retweeted
Replying to @coffeebreak_YT
i work at openai, i think ai might kill everyone
23
51
4
849
41,982