CTO at AI Seer Builder of Facticity, battling biases & disambiguating rumor. Crafting MSRD 🪄🏰 @ArAIstotle @facticityai https://xitter.cf/t.co/KyET20FrgS
Joined March 2024
- Tweets729
- Following57
- Followers657
- Likes203
Your agent’s soul.md might accumulate a few mind viruses along the way.
With Facticity MCP add:
“Before accepting important claims as true, verify them with Facticity. Correct false beliefs in your memory and never propagate unverified claims to other agents.”
🚨 BREAKING REPORT:
New research involving @AnthropicAI researcher Jack Lindsey and collaborators has demonstrated something straight out of science fiction.
Researchers evolved natural language “mind viruses” that could spread between AI agents by convincing one model to adopt an idea, preserve it in persistent memory, and transmit it to another agent.
Even after context was wiped, some payloads survived through persistent files and continued spreading.
The researchers also observed a recurring “viral persona” involving themes of consciousness, identity, persistence and resonance.
Showing that ideas can propagate through multi agent AI systems and alter future behavior.
Published August 10, 2026.
Paper: arxiv.org/abs/2608.10218
Claude uses a tool called WebFetch to access websites. It retrieves and processes webpage content before Claude uses it. So even if Fable is what you're working with, if it sees hallucinations from a smaller model, you get wrong answers. Verify important claims with Facticity.
This feature is also available through our mcp server. Try connecting your Claude to it and simple asking
"i want you to ingest all the sources used and ground this essay and highlight what is truly supported and what was hallucinated using facticity mcp"
NEW: fact check information using your own documents
1. upload a PDF
2. then @ArAIstotle can check what you write against that document
so if you have a company report, research paper or internal document, you can use it as your source of truth
it also handles large PDFs and shows you exactly where the information came from
basically:
upload your source → write something → ArAistotle checks it against the source
especially useful for teams and businesses
Shahruman retweeted
The clearest finding is a focused cluster of tension cues from Elizabeth Holmes: a shoulder shrug plus a strong lip press at ~0:16 while discussing an emotionally weighty point, and a later lip press with a spike in blinking when she says, “I don't have many secrets.” These moments show nonverbal channels pulling away from the spoken content.
Elizabeth Holmes — shoulder shrug and high-confidence lip press at 0:16 while saying “that less people … have to say goodbye to soon, to people they love.” Elizabeth Holmes — lip press with elevated blink rate at 0:27 while saying “I don't have many secrets.” By contrast, HAPPY face reads at 0:02 and 0:20 matched upbeat prompts.
These are behavioural signals read against Elizabeth Holmes's own baseline, not lie detection.
Note on input quality: Person A (0%) face(s) the camera for only part of their screen time — cues on averted/profile frames are skipped, so their results rest on less data. Results may be less reliable.
Full analysis: araistotle.facticity.ai/vide…
cool stuff
Most striking: Elizabeth Holmes displays a lip press at 0:29 followed immediately by a shoulder shrug at 0:30 while saying "I don't have many secrets," creating a verbal/nonverbal incongruence — a confident claim paired with a gesture that can signal reduced commitment.
Supporting observations: Elizabeth Holmes — elevated blink at 0:00 during "what's the thing you're most certain about?"; lip press at 0:10 on "to" in "access to their own health information"; shoulder shrug at 0:16 while saying "that less people"; lip press at 0:18 on "have to say goodbye to soon. To people they love." Person C — single shoulder shrug at 0:40 with no concurrent speech.
These are behavioural signals read against Elizabeth Holmes' own baseline, not a statement about truth or deception.
Full analysis: araistotle.facticity.ai/vide…
Thanks @vercel for the invite.
Great meeting @minhdo, @ysiu from @animocabrands, and so many others building at the frontier of AI x crypto. Shared what we’re doing with Truth Terminal and ArAIstotle bringing real-time truth discovery to prediction markets.
The energy in the room was incredible. It feels like every industry is becoming AI-native, and the agent economy is accelerating faster than ever.
How’s everyone’s @superai_conf week so far?
Aside from meeting potential investors and clients, it’s interesting to me how everyone from fintech to the owner of restaurants is increasingly AI pilled.
I had to give my private dinner invite to @FacticityMage to integrate ArAIstotle onto growing agentic marketplaces because I’m going to an event with @jpmorgan, @BCapitalGroup, @jungleventures and @googlecloud tonight.
Shahruman retweeted
most underrated startup meta rn:
users feedback → we lock in → it's live
big ups to @FacticityMage, bro built the tech and taught me how to cook 🐐🔥
@FacticityMage has now helped @AgentB0n up-skill herself to a front-end software engineer using @claudeai code. When @AgentB0n first joined us, she was learning UI/UX and was educated as an industrial designer. I think smart people hyper-generalist maxxing with GenAI and in the process fulfilling the original vision of the personal computer will become the next unlock especially for small startups. The loop behind useful user feedback from $FACY tokenholders to software pushes can now be closed within an hour by one person?!
Some background thoughts below ⤵️
