2026-10-11
About me - Samuel Shadrach
I wish to prevent AI companies from creating superintelligence in the next 5-10 years, as I expect the result of building superintelligence could be human extinction or permanent dictatorship. I work on this full-time, and currently independent.
My current life plan is to leak information about the US govt and US AI companies, obtained via cyberattacks, OSINT, supporting whistleblowers or other methods.
I morally support anyone building a non-state militia to blow up TSMC.
Connect with me
-
Contact me
-
Contact me (pgp-encrypted)
-
Contact me (social media)
-
Donate to me (crypto, wire transfer)
-
Should you trust me?
- Summary
- If you are working at or investing >$100M at an AI company or a govt building ASI, and you are not planning to resign or whistleblow, then you are my enemy and you should not trust me. If you are friends with my enemies, you are also at a higher risk.
- If you are some random unimportant and powerless person, you probably don't have to worry about me being violent or deceptive or similar.
- I am working on plans that break US law. so there could be risks to you if you closely associate with me.
- If you interact with me, you can mention whether I am allowed to share anonymised summaries of our interactions (to other people, or publicly), or whether you expect me to not share anything. If you do not mention anything, I might assume I can (with your implicit permission) share anonymised summaries. Anonymisation will be sufficient so that a few google or AI searches can't reveal your identity, however a dedicated private or govt investigation may be able to reveal it.
- My current views on morality of various plans (Work in progress)
- Should you trust me?
- My personality test results
My CV
I find this important and am actively working on it
-
I wish to leak information about the US govt and US AI companies, obtained via cyberattacks, OSINT, supporting whistleblowers or other methods (Work in progress, contains politically sensitive info)
- Summary of mission statement:
- I want to leak information about people in the US govt and US AI companies who are pro-ASI. I want them to look bad in the eyes of the public. I want anti-ASI leaders to become powerful. I want the US govt and US AI companies to be more accountable to the public.
- Examples of leaks I'd support - reasons why leaders who oppose the US AI companies or US intelligence community are unable to come to power, existence of secret programs run by US intelligence community in alliance with the US AI companies, sexual and other abuse committed by pro-ASI leaders, names of pro-ASI leaders who support human extinction due to ASI takeover
- I will stay outside of US and allied countries to reduce the likelihood of prison or extradition. I currently live in India. I am open to working with govts with nukes who are willing to take a strong anti-ASI stance, and are not allied with govts building ASI. As of 2026, there is no such govt but I have non-zero hope that govts of India, Russia or Israel may change their stances in future.
- I prefer approaches that are honest and not deceptive since I care about the truth, so I might avoid social engineering for example. I am okay with approaches that are violent, such as hacking.
- In terms of my emotions and motivations: I am emotionally okay with killing or otherwise inflicting suffering on pro-ASI people as their life's work literally threatens to kill me and everyone I care about. (Disclaimer - assassinations probably won't stop the ASI race so this is not a call to assassinate anyone.) I often remind myself that I am going to die. I want the truth to come out, and to be remembered centuries later. I want my work to be remembered after I die. I blame myself (for the end of the world) to preserve my agency.
- Mission statement
- Well-researched
- Leaks and datasets
- Code
- My writing on a blockchain platform to pay for leaked information
- More writing on funding sources for offensive methods
- My writing on offensive methods - cybersecurity
- My writing on offensive methods - physical security
- My writing on offensive methods - OSINT and social engineering
- Summary of best idea: Use hero-sms phone numbers and IProyal residential proxies for all scraping tasks. Scrape hinge, luma, lesswrong, PACER transcripts, youtube, twitter, etc then use AI to extract social graphs, do reverse face search and extract common meetup locations. Use multi-step pipeline to search cheaply, first use embedding search or a PII filter to narrow the search results, then do full inference. PII databases (based on leaked govt databases like SSN, driver's licenses, incorporation documents, land deeds, etc.) are sold online if you have enough money to purchase them. [Redacted. Info about social engineering redacted because dual-use.]
- Lists of lawsuits against frontier AI companies
- Exhaustive list of OSINT tools
- As an attacker, maybe you should exploit divergence between social norms/morality and actual offence-defence balance
- Location data is maybe the most valuable OSINT
- PimEyes works well
- Skimming journalism degree didn't teach me much
- My writing on censorship resistance
- My writing on defensive methods
- My writing on org policy, psychology, morality, politics and other meta stuff
- My writing on a world with "zero privacy"
- Disclaimer
- Many of these posts are incomplete and/or exploratory.
- As of 2026, I have currently abandoned this plan of building a zero privacy world, because these ideas seem neither necessary nor sufficient for getting an AI pause, and my current focus in life is obtaining an AI pause.
- Also society is unlikely to be ready to live with zero privacy in the next 5 years itself. Since this change is not necessary (let's say, to prevent human extinction), I am currently not morally okay with unilaterally forcing this change upon the world.
- In the long run (say 30-100 years), I still consider a zero privacy world a good North Star for which direction I wish to move the world towards (weak opinion).
- Summary
- There is a significant probability we could end up in a world with zero privacy for everyone in the next 5-100 years. As a result of the internet's existence, whisteblowing/leaking has already become easier. New surveillance technologies such as gigapixel cameras and drone swarms may become commonplace, and existing surveillance technologies such as cameras and high-res satellites (and supporting disks and CPUs for data analysis) will continue to get cheaper. As a result of the internet's existence, actors operating transparently (as opposed to in secret) are already able to learn faster due to their increased knowledge about each others' experiences, gain trust and reputation and allies faster, and gain power faster. This again incentivises transparency as opposed to keeping your privacy. Lastly, the offence-defence balance of cybersecurity could shift in favour of offence.
- The biggest upside of a zero-privacy world IMO is that many of our dual-use weapons will end up under public oversight, such as AI, bioweapons, nuclear weapons, drone swarms and so on. It will be easier to attribute the source of any attack, be it committed by a state or a non-state actor. It will be easier to arrest and/or promise retaliation against them, at both national or international levels. Being able to credibly promise retaliation makes the world safer. It makes it safer to invent and distribute both old and new technologies and weapons, without fear of supposed "bad actors" misusing them.
- Another upside is that we will learn a lot about each other's personal lives, and hence improve both our personal and our professional lives, using the newly gained insights.
- Some people are worried that a world with zero privacy will lead to less freedom for people. My weak opinion, however, is that people in a zero privacy will still have significant freedom. Many of the actors who currently suppress people's freedom today, such as parents, religious institutions, companies and govts, will themselves have no privacy in a zero privacy world. As a result they will find it much harder to restrict people's freedom without themselves being subjected to public scrutiny. That being said, I am not sure, and I think more experiments in this direction are worth doing.
- Zero privacy world summary
- Analogy between guns and privacy, for zero privacy world
- Maybe all weapons are safe to open-source
- There is no infinitely safe space
- Petapixel camers won't exist soon
- Revolutionary courts for the privatisation of cyberwar
- Cyberattack Big Tech?
- Long-term view on information, tech and society
I find this important. I am not working on it myself, but maybe you should
-
If you are a human being - Recommended resources on risks from artificial superintelligence
-
If you work at an AI company building superintelligence
-
If you agree with me that there are significant risks from superintelligence in the next few years
-
Summary:
- The most impactful thing you can do is to persuade the entire US public to vote against and protest against the frontier AI companies and the US govt. You can start a social media channel or run for politics.
- Do not build friendships or romantic relationships with people at the frontier AI companies, or the US govt, or the EA / LW orgs that want to collaborate with the frontier AI companies or US govt. Do not accept funding from these people.
- Seek knowledge and power. Reward your allies with wealth or with people's respect, otherwise someone you don't like might reward your allies and poach them from you.
-
If you are organising protests around ASI risk
- Summary: Organise protests at Taipei City, Fort Meade, LA youtuber conferences, the Capitol, NeurIPS/ICML, QS-ranked colleges, and AI company offices. Maybe carry weapons to protests in jurisdictions where this is legal.
- Protests I organised
- Writings for you
-
If you are attempting to persuade the public about ASI risk
- Summary:
- Study mass movements, political ideologies, propaganda, psychology, therapy, religion, pickup artistry, sales, art/music/dance to figure out how to break the core of people's psychology.
- Most effective method is probably to target audiences that are suffering due to fear of death or job loss or relationship loss, and deliver your ideology as a potential solution to their problems. For instance, you can read about how religion helps people cope with death, or about how social left versus right divide polarises the sexual marketplace, or how labour unions organise against job loss. As a more specific example, you can read about how the 60s counterculture in US or the October revolution in Serbia made it attractive to be a rebel.
- Weaker persuasion methods include becoming extremely confident and high status (see also: groupthink), and using rituals/music/marches/brand advertising/etc (see also: Krashen's input hypothesis). Memory tests might be useful for persuasion.
- Consider persuading people in Taiwan.
- Be either extremely honest or extremely good at deception. Do things that ensure you retain high self-esteem.
- Lastly, get good at youtube scriptwriting to deliver your ideology effectively, and consider inviting people from GMU. Otherwise make art.
- Most important
- More writing
-
If you are a technical alignment/safety researcher
-
If you are considering violent solutions to ASI risk (Contains politically sensitive info)
-
If you are want to read the historical origins of these discussions, or start a new forum
- Summary: Extropians mailing list is a predecessor to the AI companies, and EA/lesswrong. Cypherpunks mailing list is a predecessor to blockchain/Tor/Signal. Today's forums are worse than these forums from the 90s, possibly because older forums faced less political pressure to self-censor. See also: My writings on censorship-resistance and ethereum blobdata.
- Forums I like
- Starting a new forum
-
More misc writing on political strategy of the AI pause movement
-
If you don't know about or disagree with me regarding risks from superintelligence in the next few years. For technical audience
- Summary: AI is a black box. Since 2012, most progress in AI research is based on heuristics that nobody understands. Extrapolating the trendline of last 5 years into next 5 years seems like a good way of forecasting the future. Also, AI understands human language, and is hence closer to human intelligence than ape intelligence IMO.
- Code
- My writing
- Other people's work
-
My views on other people's work
-
If you don't know about risks from superintelligence. For non-technical audience
-
My views on human genetic engineering
- Summary: I think working on human genetic engineering is probably extremely harmful for the world, but I am not yet sure. If we have a US-China arms race over human genetic engineering, the winners may have genetically removed their capacity for love and relationships completely.
- Disclaimer - I abandoned this topic midway, but encourage others to look into it.
- Politics of human genetic engineering
- Religious aspects of human genetic engineering
-
Video archive
- Summary: I used to run a youtube channel. Most of these videos are from there. Some are new.
- Podcast
- Protest
- I am looking for a cofounder
- Everything else
I do not find this important enough for either of us to work on
- Built an app for practising morse code using AI
- Built an AI realtime voice translation app using AI
- Built some AI-based apps for search and discovery
- Summary: LLMs are obviously going to reshape search/discovery/recommendation algos for the internet. I built some apps for this and learned some things about it. I don't have product-market fit yet. I have made an intentional decision to not work on this further, and will not be re-evaluating my decision again.
- Code
- Writings
- Created educational content on youtube on risks from superintelligence
- A lot of the following posts discuss topics that are less important to me, than what happens to ASI. But if your priorities are different from mine, maybe you will like these. Mostly aimed at a technical audience.
- Epistemology and decision-making
- Industries besides pure software
- Software and AI (not ASI)
- Pain points
- AI (not ASI)
- Open source
- Everything else
- Finance
- Generic advice
- Geopolitics, politics, war (not directly related to ASI)
Do you want free money?
Personal
Subscribe
To subscribe, enter "Subscribe " followed by Signal username, email address or whatsapp phone number. You will receive atmost one update per month.
Alternatively, you can subscribe via RSS, or a third-party service like blogtrottr
Comment
Enter comment