Home | Search

Samuel Shadrach newsletter (issued 2026-06-29)

Hi everyone

If you are new to this newsletter, I strongly recommend starting from the following posts

• I am looking for a cofounder to cyberattack the US govt and US AI companies, and leak information about them - I will pay you $1000 if you refer me to such a person
• Discussion forums for AI extinction risk
• Reading list on persuasion for anti-ASI movement
• How to support the movement against risks from superintelligence
• You can also browse through my website and figure out what appeals most to you

Concrete work I did over the last two months

• Audio for Elon Musk x Sam Altman lawsuit - I scraped the livestreamed audio of the lawsuit. Sharing this is illegal as per US law, but I am sharing it anyway
• I wish to give a journalist my scrapes of social graphs of US AI companies obtained from twitter and youtube - Scraped social media and used AI to extract social graphs of who knows who. Possibly useful for investigative journalists and OSINT researchers
• Polymarket question "Will Sam Altman go to jail" can be used to incentivise insiders who can put altman in prison - You can invest your money here, to create incentives for whistleblowers. You can also tag polymarket on twitter and ask them to open more questions like this.
• My protest at India AI Impact Summit 2026 - Updated to include videos of the protest
• Video: Samuel x Raghav - Is research taste a bottleneck to ASI?
• I have decided I will be working on leaking information about those building ASI, atleast for the next few months and probably even after that. I will not be working on persuading people about ASI risk. Persuasion is more impactful than leaking information. But I dislike talking to people so I think a full-time role involving persuasion is not what I'm suited for.
• I wrote more blog posts, linked below

Some additional important posts in the last two months

• Important, about leaking information
  • List of OSINT and cyber startups that work with the Indian govt - I am semi-seriously considering raising funding via this route. I probably need to find a cofounder first.
  • Polymarket users have paid for leaked information
  • Govts have paid spies using cryptocurrency
  • Blockchain distributed oracles for crowdfunded bounties for leaking information - A missing piece in the infra to pay a crowdfunded bounty to whistleblowers, as incentive for their leaks
  • A startup to enable ETH-XMR atomic swaps may be profitable and help whistleblowers as a side effect - A missing piece in the infra that could let whistleblowers publish their leaks directly without relying on journalists
  • PimEyes works well - Reverse face search tool
• Important, about ASI risk
  • AI timelines as of 2026-06-18 - I wrote yet another post on why I think ASI might get built by 2030 (unless there is a political movement to stop this). Incorporated a lot of useful feedback I have received from people.
  • Persuasion
    • Maybe especially focus on persuading Taiwanese govt and public about AI pause - Taiwan is an independent country and a democracy, and can unilaterally destroy their semicondutor fabs without the permission of either the US or Chinese govts. Maybe persuade them directly about ASI risk
    • You can use virtual art galleries to inspire people - I made a virtual art gallery
    • Maybe anti-ASI protests should be done as tours - Suggested possibly high-impact protest spots including Fort Meade, Taipei city, youtuber conferences in Los Angeles, NeurIPS and ICML conferences, and outside QS-ranked engineering colleges.
    • If you study at a top 1000 QS-ranked university and have 1000 social media followers, I can pay you $100/mo to run ads of famous anti-ASI youtubers on your social media account
    • Laundry list of ideas on public persuasion around ASI risk
  • Censorship-resistant forums
    • Moderate forum on ASI risk outside US geopolitical sphere - If someone is willing to become a moderator full-time, I am open to hosting an online forum on ASI risk, that can contain information critical of US govt, while I live in India.
    • Why were cypherpunks and extropians better than today's social media? - Important for anyone building a new forum
    • My views on lesswrong - I am almost (but not yet) banned from lesswrong
  • Morality and coordination
    • My actual views on assassinations to prevent ASI - I have no moral issue with assassinations. However, I don't think assassinations are an effective strategy. Many people in EA/LW circles do not tolerate this opinion of mine, especially in public forums, and especially those who are themselves funded by or friends with people at US AI companies or US govt.
    • Slides on morality - I am trying to extract generalisable insights around morality
    • Long list of mental models on morality
    • It is good if your ingroup has stable emotional generators - Building stable ingroups and ideologies is hard. Focus on the emotional generator of the opinions not just the technical details of the opinions
  • Some conversations I've had with people on risks from superintelligence

Maybe less important posts

• Maybe less important, about leaking information
  • What happened to Vincent Canfield (cockbox) versus Michael Hirdes (CCC) case - Vincent Canfield operates a censorship-resistant VPS and email service called cockbox. He accused people at CCC, the famous left-leaning hacker conference, for beating him up over political differences. He has now shifted to eastern Europe and migrated his servers there. I want to know what happened to the lawsuit.
  • I suspect prices of residential proxy providers can be undercut - Useful for bulk scraping social media
  • Skimming journalism degree didn't teach me much - Journalism involves skills I lack, but my guess is these are learned via practice not theory
  • Jailbroken models can delegate subtasks to non-jailbroken models - Jailbroken AI models are useful for OSINT and cybersecurity tasks.
• Maybe less important, about ASI risk
  • How I hired a video editor - Possibly useful if you are newly starting to make video content
  • AI pause cold emailing infra needs a lot of improvement - If you are going to insist on bulk spamming emails at influential people, atleast build good infra for it
  • My views on Midas project - I appreciate them for not working under any of the AI companies
  • My views on Visakan Veerswamy - I respect Visakan's views a lot. However, Visakan's views seem to imply it is never a good idea to build a political movement against an oppressive govt, or wage a war against an enemy. This has obvious counterexamples in my view.
  • My views on Chris Lakin - Similar-ish criticism as with Visakan
  • Core of memetics versus scaling what works - My bad attempt at comparing a lot of different ways of spreading your ideas such as social media ads, in-person protests, in-person banner ads, and so on.
  • Is it true that AI company CEOs without children are the most reckless - Attempted to evaluate socially right-leaning hypothesis that having children makes you more altruistic, in the case of ASI risk
• Maybe less important, about other topics
  • Most important geopolitical graph - world energy balance of trade - Who owns nuclear weapons is probably the most important factor driving geopolitics. The world energy balance of trade is probably the second most important factor
  • Your file format should have a simple parser in C or Rust - Most file formats don't survive multiple decades, here is my guess on how to ensure yours lasts
  • Git sucks with big files - I thought I am a bad developer but turns out git actually sucks with big files
  • Run SaaS on user-controlled servers - Maybe this is a viable business model that lets non-technical users pay for a managed solution, but without sacrificing freedom or control.

Subscribe

Enter email or phone number to subscribe. You will receive atmost one update per month.

Alternatively, you can subscribe via RSS, or a third-party service like blogtrottr

Comment

Enter comment