Yeesh. I guess that makes one of us. Truly one of the more horrifying short stories I’ve read.
> I put a band around my head, and open my laptop at Conduit. It’s 9 am. My device pairs with my laptop over Bluetooth. I open, without loss of generality, Codex.
At least we are reassured that short-range networking protocols will be invariant under the apocalypse.
It's been explained to me by neuroscientists I respect that EEGs and friends are like trying to listen in on a conversation happening on center ice from the parking lot of the arena. You can maybe tell if there is a match going on if there is a rowdy crowd but that's about it.
>Thought-to-text, aka telepathy
No, that's not telepathy. I'd leave the term alone. Consider it to be reserved for... well... telepathy!
1. Superhuman AI is inevitable.
2. Writing will become a bottleneck to communicate effectively with it.
3. Higher communication bandwidth would let us keep up with the machines.
4. Therefore with higher bandwidth humans could stay part of AI decision processes.
5. That makes it more “human”, as humans would be kept in-the-loop without becoming a bottleneck to be worked around.
6. If you let superintelligence write into your brain, that lets you merge with the machine.
What a terrible idea.
Technically fMRI is noninvasive; it is just a bit inconvenient for everyday use.
Sounds magical
Ah, finally! To be liberated from the burden of speaking! Perhaps soon the AI will even think for us!
On a more serious note, this is additional evidence that Wall-E stands as the most accurate sci-fi projection of the future.
- still annoyed by AI-speak
What a bold vision!
"I'm leaving OpenAI to build telepathy"
I feel new a trope is being born.
I read "race conditions" all over this vision.
Ah, yes, very joyful indeed.
Also, fuck those interns! They're just too smart to keep up with.
"I am a former-OpenAI employee and looking for investors to fund my AI startup."
When we have GPT-7 it will still do AI-speak?!
I'm trying to remember if I was as much a tech-optimist as she is, at her age. I remember fondly imagining a futuristic smart PDA (Personal Digital Assistant) that's everywhere with me and I can use it like in Sci-Fi movies and games.
I'm assuming this is roughly the same vision, for the next generation that already has smartphones?
Yet something about this new "modality" of listening in on your thoughts, hearing your brain, is just so icky and dystopian to me. Guess I'm old?
I cannot comprehend the output from AI fast enough for the speed difference between a verbal and thought response to matter.
Not to mention the privacy nightmare of logging my raw thoughts in a corporate environment and potentially routing them to an LLM provider.
But sure, you're right. Reading is not a deterministic mapping of text to thoughts.
Apart from the trouble of defining what a thought is, independent from the person who thinks it.
Me thinking of grabbing a mug of coffee is a different thought from you doing the same... defining "thought" pulls in a whole jungle of epistemology (maybe I should say philosophy, more broadly, I'm not in the field).
Is there an abstract concept of "the thought of raising your left index finger"? Or "the thought of remembering severe pain"? Etc etc
The only thing I would actually want this for would be a dream logger. Extremely invasive and quite dangerous otherwise.
How come these brilliant machines can’t figure out how to make a hospital stay not cost $60,000? Or build a car for less than $30,000? Or keep a megalomaniac from fooling enough people to seize power in a country of 300 million people?
Under current US law, can people entering the US be subjected to mind-reading?
“A couple years” Not “a few years” — “joyful extension” — that’s an optimistic prediction in two ways…
In a couple years?! Get over the technical hurdles, do you think PEOPLE will be ready to telepathically talk to a computer in A COUPLE OF YEARS?
Can this technology also do text to thought?
How much are those people making to be able to afford things like this?
He starts off at 2027, where he’ll have a headband that can read his thoughts. That’s in 4-16 months. Correct me if I’m wrong, but last I checked we‘re quite far from being able to read people’s thought from a headband.
Do these clowns actually forget that part?
ever since Elon did his "promise full self-driving every year" thing, entrepreneurs been on a weird sort of fiction escalator.
thought-to-text is incredible. imagine doing something that remarkable with your life and feeling the need to exaggerate.
But, obviously those only make up a small number of people.
Strapping a mind-reading machine onto some detainee is probably the more likely case.
Poe's Law at work.
The problem is that anything based on LLM architecture is not going to be fast or accurate enough to allow someone to, say, fly a fighter jet with this interface.
No, they are aware. You and I are not the audience for this post. People with capital to invest are the intended audience.
It is been 10 years ago when I looked through some research in that area. Back then it was the problem to get enough good signal. Has the situation changed radically here for non-invasive electrodes or portable detectors?
I understand that the approach would be to turn weak signal, context, and LLM based signal processing into useful human computer interaction. For everything else, I have the feeling we are still in speculative territory.
The statement implies that the company has some secret hardware that can, or at least, is close to doing that. That would blow my mind, but the author seems to believe it.
As an aside, I believe the author is a she.
Neuralink exists since 10 years, is in clinical trial, has been implanted on a living human being in 2 years ago.
The lack of adults in the room to set rules around moral hazard is hardly her generation's fault.
This is a Black Mirror episode coming to life.
People like this are the input side of the "oops we built the Torment Nexus" pipeline.
The Conduit® Thinkerface® Pro Band will contain a small 90B dense local model that pre-screens your Conduit® NeuroTokens® stream for compliance against the Conduit® Acceptable Usage Policy applicable for your region and subscription tier.
If non-compliance is detected you'll be promptly notified via a small current applied to the Conduit® ThinkRight® Human Training Electrodes embedded within your band. Intensity and duration are dynamically adjusted until compliance is restored.
Even better, because this all happens on-device before your Conduit® NeuroTokens® are sent to the Conduit® NeuroServer®, non-compliant thoughts don't count towards your Conduit® Daily Thinking Allowance or expend credits from your Conduit® Additional Thinking Purchase Plan.
So you'd need to force people to somehow think the same things they do when talking or typing, without actually performing those actions. If you can do that, why not just have them talk?
> It’s hard to collect enough data using invasive methods. Few people want a chip in their brain! But non-invasive methods are getting much better. The hardware is improving and getting cheaper, though I apologize for being vague about the particularities of our hardware.
It doesn’t even really need to work of course.
What collapse of web 2.0?
Technically the website we're on is web 2.0 because we're interacting as users.
I don't think it's naiveté on their part to want to build new things. I think you're just jaded and hyperfocusing on the negatives.
Unfortunately I think this is inevitable because as a species we perform a steady march toward collective desires for items of science fiction, so start your adversarial double-think practice now.
And this will provoke “well we need to regulate AI then”. No, you need to regulate poking around in people’s brains without asking them before that becomes normal.
https://condu.it/thought/10k-hours
Their data acquisition routine was driven by Claude and they were recruiting random people off Craigslist pretty aggressively a few months back. I know a couple of folks who responded out of curiosity and described a very janky setup in the rented basements of a couple of residential houses in SF. From what I heard about the setup the data they got must have been incredibly noisy.
To the people who are sceptical here, I wouldn't count them out yet. They almost certainly can get funding with what they captured, which would get them better hardware. And given the state of brain scanning - people have captured images from scanned brain activity some years ago as I recall - I'd say they have a good shot. But yeah, it's not telepathy.
Previous HN discussion:
We know that we can use classical techniques, (non-AI), to psychologically condition humans. In the lightest case, we do it with advertising and social media. I'm absolutely certain that state-level adversaries use even more effective techniques to train and condition agents for fieldwork.
If your adversary knows you are more likely to rely on intelligence gathered in this fashion due to its reliability, they can definitely game that. Or rather, I guess, I would definitely game it if I knew the Russians had this kind of technology. So I don't know why they wouldn't be thinking along the same lines.
Basically, make them think their new technology is working.. until it counts. If it consistently works 99 out of 99 times. They'll have no reason to believe it's not working the 100th time.
But I guess if you were interrogating, maybe a pot dealer or spouse muderer or cheating husband or something, it might be valuable.
Talking to a machine has the same effect on me.
When I write something, I deliberately curate my thoughts in order to convey an idea, often heavily depending on the context. Context encompasses the whole slice of the world that is connected to my individual life, who I want to address, and what I want to convey.
As a simpler counterexample: it's possible to write down lies, deceptively. The definition of a lie means that I know better than what I say.
So, "honest thought" != "fact":
where does this take us?
People who lie not only try to deceive others about facts, the core of most lies is deceiving about intent.
I used the word "idea" here, and I think it is nothing more than a rephrasing of Platonic ideas, but I ought to be careful, because we are more than two thousand years in the future from there.
August 4, 2026
On Thursday, July 23rd, I resigned from OpenAI. On the 24th, I started as a Founding Researcher at Conduit. We're building telepathy: thought-to-text models, trained on non-invasive neural data.
I'll talk about:
My prediction: In a couple years, the main way we'll talk with our AIs is with our thoughts. They will not be a super-smart automated intern or coworker that you desperately try to keep up with. They will be a natural, joyful extension of you.
I wrote down vignettes of this future below. I view them as optimistic but highly plausible.
2027 — A morning at Conduit
I put a band around my head, and open my laptop at Conduit. It’s 9 am. My device pairs with my laptop over Bluetooth. I open, without loss of generality, Codex.
I think about the day's work as I look at my code, review yesterday's notes, and prepare for our 9:30 am standup. My GPT-7 agent has been chugging away at the new encoder I’ve been exploring for over a day now. I look at its work. I become confused by the plot labels, annoyed at the AI-speak in its first paragraph, and curious about the symbols in the equations. My vague thoughts get sent to Conduit's model, which uses its priors over language and the kinds of things I might say to output:
conduit://thought-stream → codex
$ decode --source=neural-latent --autosend --to=codex
I’m looking at your plots. Please relabel the legend, and add a text cell below with a simple description of each line.
I understand your 2nd paragraph of discussion, but please rewrite the first paragraph using the humanizer skill.
Make a textbox explaining the equations with cleaner notation.
The plot, paragraphs, and notation each took me 10 seconds to glance over. I have auto-send on, so after each 10-second chunk, the Conduit model deciphers my thoughts and sends it to GPT-7, which assigns each incoming task to a subagent. I look at the new plot, paragraph, and equations. I wonder whether spherical harmonics might be useful after all; Codex spins off a subagent. Convinced by the plot, I decide to make a quick slideshow with a prettified version showing only the baseline and top two lines, to show my teammates; Codex spins off a subagent.
I make myself coffee. My thoughts are mostly empty, but I briefly recall what I want to say to the candidate I’m getting lunch with tomorrow. In the background, Codex is prompted to think about info I might want for the lunch chat, and decides that updating the synthetic data scaling plot before showing it off would be prudent; Codex spins off a subagent.
( (
) )
.--------. | |] | COFFEE | | | '--------' \____/
I'm not saying words really loudly in my head while getting coffee. I just read the plots as I normally do, and make coffee as I normally do. It feels like magic.
2030 — Industry adapts and Conduit expands
The AI companies now train their models to directly interface with Conduit’s latent representations. My encoded thoughts get sent directly to Codex, rather than having to pass through Conduit's decoder model first. That means I easily communicate thoughts that are hard to describe in text, like mental images.
I rarely go band-less when chatting with AIs these days. It's annoying, honestly, to go without. With the neural headband, I feel like I have superhuman powers over my laptop! Without it, I feel like I'm talking with a superpowered alien who's trying its best to be helpful but isn't sure what I want and is scared I'll get mad if it does something I don't want. Ugh.
Conduit continues to iterate on non-invasive read, but we’ve spun up two new efforts: invasive general read, and general write.
Most people are happy to stick with their neural bands, but a good number are excited to get higher fidelity reads via invasive tech.
More interesting is the recent excitement in writes. I feel superhuman in my control over my laptop. But my senses are still merely human. It's like if I could control my arms, but I'd lost all feeling in them. Yeah, I can still see my arms, and it's way better than not having arms, but it's still really odd. Like Ian Waterman. I want to feel what my Codex feels. Now that Conduit does general read, learning to do general write is many OOMs more data efficient.
More importantly, I want to unlock the other applications of write technology. I want to make my brain as efficient and neuroplastic as when I was 18.
2035 — The AI is no longer “other”
My AI is a natural extension of me. It feels like a sixth sense and another limb. I wonder about a problem, feel as though I’m literally surfing the web, see glimpses of the websites, get flashes of intuition about the problem, and ultimately derive the answer. It feels fun! My brain is like the Flash. Because of write tech, my thinking is the fastest it's ever been even when I turn my AI off.
How superhuman 2035 looks depends directly on how superhuman we, as society, decide to make our AIs. Perhaps we choose to pace ourselves. But I like that this is literally a human-in-the-loop vision of the future, where AI directly empowers humans rather than replacing us.
A few quotes about the future of thought-to-AI
Writers and speakers more elegant than I have elaborated on this thought-to-AI interface over the past decade. I will link to them, along with brief excerpts.
In theory, it’s simple. Our input is brain activity, and our target output is what the person was doing at the time – for example, what text the person wrote. Given the brain activity, we want to predict output that is semantically similar to what the person wrote.
To train models that can predict text given brain signals, we must apply the same lesson learned by those predicting text given speech audio, or text given preceding text: the bitter lesson. The lesson roughly states that you should throw more useful compute at your model, and your model will become better than any ingenious algorithm you could've hand-crafted. That means we must scale up our data collection by orders of magnitude beyond what has ever been done in academia.
It’s hard to collect enough data using invasive methods. Few people want a chip in their brain! But non-invasive methods are getting much better. The hardware is improving and getting cheaper, though I apologize for being vague about the particularities of our hardware.
As we're training on more data, the model is predicting text that is more semantically similar to the subject-written text. Yes, there's some irreducible error due to noise, but for most modalities we're not yet in a regime where we're pushing against that. Concretely, the scaling laws are looking good: the cosine similarity of our latent space predictions with the target latent spaces goes up as a straight line with respect to the logarithm of the number of hours of data. We're in the GPT-2 era.
We don’t need perfect decoding to be useful. Your thoughts will be like GPS in a city: a noisy GPS signal isn’t enough to determine your exact location. But combined with a map and a navigation route — equivalently, the LLM and context — it becomes remarkably accurate.
To read up on how we do data collection, check out Conduit's blog post. If you're more of the active learning type, come be a research participant!
3 reasons:
In short: I’m happy! I’m working on a problem I’m obsessed with, that is frighteningly ambitious, with a small group of people I like.
For those who don’t know me: hello! I left OpenAI two weeks ago, after spending 1.5 years there as a researcher.
I grew up in Washington State, where I played competitive chess from ages 5–15 and stopped after becoming a WIM. My first time hearing about the potential for smarter-than-human AI was when I was 12, but for years I figured it was just a weird but interesting idea that people on the internet liked to write about.
At 18, I enrolled in Harvard to study computer science. In a class, I learned about GPT-3 and finally read Bostrom’s Superintelligence. Wow, what a wakeup moment. I got invested in AI safety research, had a brief stint in AI policy, and ultimately joined the alignment team at OpenAI. At OpenAI, other than my research, I spent some time on various side projects, including making OpenAI’s AGI onboarding presentation and our alignment blog, and helping advise the AI Resilience division of the OpenAI Foundation.
If you want to chat about Conduit, reach out to me at naomi@condu.it. Let’s grab coffee, or I’ll give you a tour of our unusual and beautiful office in San Francisco. We’re always hiring researchers, infra folks, and operators.
Acknowledgements: Thank you to Aidan Smith, Devansh Pandey, Julia Shephard, Lev Chizhov, Ryan Kaufman, and the Conduit team for feedback. I'm grateful, too, for the several other companies building ambitious BCI tech, including but not limited to those building better sleep, better ultrasound, a solution for blindness, and uploads - I'm excited to use your tech! All mistakes are my own.
As for Utopias/Dystopias, I don't know how to answer that without waxing poetic about ideas like chaos being the natural order, or beauty being in the eye of the beholder, <insert other platitude>.
I’m generally pro-privacy, but here’s the hard truth: at my startup, we disabled keyboard input in our app Slopgen and switched the default to Conduit Thought Input™, and only 2% of users bothered to navigate the 17 submenus and submit their SSN and face scan to re-enable typing. Yep, 2%, that’s all.
DAUs only decreased by ≈11%, but our ARR is already up 170% thanks to our AdThink™ revenue sharing agreement, and only 1% of users argued with our Fin.ai bot for the 45 minutes needed give a thumbs down feedback about the change.
Let’s face it: talk is cheap but users *want* these features, revealed preference *always* tells the real story. Better get used to it now or get left behind.Again look at the run around on the fourth amendment or the commerce clause ruling that made it so growing wheat for your own consumption was legally considered interstate commerce.
And, my point was, there may be zero words to collect at all, because they'll tend not to be actively thinking about any words.
https://en.wikipedia.org/wiki/Kyllo_v._United_States - The case I was referencing.
The fourth is effectively gone in the digital age since so much activity is done via third parties who the government can compel to hand over your information.
And for the interstate commerce clause they twisted it to argue that completely _intrastate_ commerce was actually interstate commerce because you would have bought the goods from other states if you didn’t make it yourself[1].
I’m not even opposed to the ruling in terms of what the government should be doing, but you are being blind if you think that’s logically consistent. There’s no reason the constitution would specify “…to regulate Commerce with foreign Nations, and among the several States, and with the Indian Tribes.”
It would have just said they had the power to regulate commerce.