Latest Corrections
X Feb 28, 2026 at 07:29 AM
x.com/SandyofCthulhu/status/2027164167233232920
- This is an overbroad universal claim. African wild dogs are successful predators, yet credible references describe them chasing prey for kilometers and sometimes beginning to consume/disembowel prey while it is still alive, not immediately killing or immobilizing it or avoiding close contact.
X Feb 28, 2026 at 07:21 AM
x.com/ryancbriggs/status/2027097790371766592
- The underlying UK Biobank accelerometry study reports ~31–32% lower incidence for *physical activity–related* cancers, not for cancer overall; for total cancer incidence at 4.5 min/day, the association is ~20% lower.
X Feb 28, 2026 at 07:21 AM
x.com/amypretzel/status/2027064869040931329
- The “80 years of a working, functional” zipper is incorrect: early zipper-like designs were unreliable, and the modern practical zipper dates to 1913 (patented 1917), shrinking the invention-to-adoption gap to decades, not 80 years.
Substack Feb 28, 2026 at 05:41 AM
graymirror.substack.com/p/redpilling-claude
- “Woke” was being used as a pejorative by conservatives/Republicans well before the 2022 midterm elections (at least by the 2010s, and clearly by 2019).
X Feb 28, 2026 at 01:31 AM
x.com/twitter/status/2027555979563245752
- Multiple news reports quoting senior U.S. officials say Anthropic raised concerns about whether Claude was used in the Maduro raid, implying potential disapproval—i.e., there *was* public indication of a complaint/concern.
Substack Feb 28, 2026 at 01:27 AM
thezvi.substack.com/p/anthropic-and-the-department-of-war
- DoD Directive 3000.09 is a DoD policy directive and does not make conduct "illegal" in the sense of creating a law. It also does not categorically require a "human in the kill chain" for all kinetic/autonomous weapon uses; it calls for "appropriate levels of human judgment," which DoD/CRS notes is flexible and not the same as manual human control.
X Feb 28, 2026 at 12:46 AM
x.com/twitter/status/2027542236263514490
- Federal suspension/debarment generally excludes a contractor from federal contracts (and certain federal subcontracts/covered transactions), but it does not create a blanket legal ban on doing business with any private company that happens to do business with the federal government.
X Feb 28, 2026 at 12:42 AM
x.com/ThePrimalDino/status/2027414769372676410
- NASA’s Gateway program is not dependent on the SLS Exploration Upper Stage (EUS): Gateway’s foundational elements are planned to launch on SpaceX’s Falcon Heavy, and NASA describes Gateway as involving multiple launches on different vehicles.
X Feb 28, 2026 at 12:21 AM
x.com/robbensinger/status/2027533660048789868
- Anthropic did not “drop all of its safety commitments” that week; as of Feb 24, 2026 it publicly maintained and added multiple safety commitments (e.g., publishing a Frontier Safety Roadmap and model Risk Reports) under its updated Responsible Scaling Policy framework.
X Feb 27, 2026 at 09:23 AM
x.com/daniel_houck/status/2027204733111349464
- OpenAI’s ChatGPT was introduced on November 30, 2022, while internet-browsing features rolled out months later (spring 2023). So it’s not accurate that connecting it to the internet was “the very first thing” they did.
LessWrong Feb 27, 2026 at 09:17 AM
www.lesswrong.com/posts/d5Lqf8nSxm6RpmmnA/anthropic-statemen...
- “Supply chain risk” is not a label reserved for U.S. adversaries; in U.S. procurement law it is defined broadly to include risks posed by “any person,” not just adversary nations or their firms.
X Feb 27, 2026 at 08:36 AM
x.com/tyler_m_john/status/2027074281499431306/photo/1
- The name Darius (and thus Dario) is generally etymologized as meaning “possessing goodness” or “holding firm the good,” not “Lord.”
Substack Feb 27, 2026 at 04:28 AM
thezvi.substack.com/p/gpt-4o-is-an-absurd-sycophant
- The post misspells Shopify’s CTO’s surname: it is Mikhail **Parakhin**, not “Parakin.”
Substack Feb 27, 2026 at 03:58 AM
thezvi.substack.com/p/dating-roundup-11-going-too-meta?utm_s...
- U.S. government statistics indicate divorce (or marital disruption that largely becomes divorce) is well above 30% for first marriages, so stating the lifetime divorce rate is “only 30%” is not accurate as a general claim.
Substack Feb 27, 2026 at 03:55 AM
thezvi.substack.com/p/one-true-love?utm_source=publication-s...
- The Substack post displays a publication date of Feb 09, 2024, not August 29, 2023.
LessWrong Feb 27, 2026 at 03:46 AM
www.lesswrong.com/posts/DMswzhPQqkqx2XAma/a-full-epistemic-s...
- The famous “customer complaint” tablet the post cites is Old Babylonian and written in Akkadian, so attributing it to “the Sumerians” is inaccurate.
Substack Feb 27, 2026 at 03:42 AM
read.technically.dev/p/vibe-coding-and-the-maker-movement
- Anderson did leave WIRED to run 3D Robotics, but 3D Robotics already existed—he co-founded it in 2009, years before he stepped down at WIRED in 2012.
Substack Feb 27, 2026 at 03:37 AM
thezvi.substack.com/p/openai-moves-to-complete-potentially
- Patagonia’s own announcement says only profits not reinvested in the business are paid out as dividends to its climate-focused nonprofit, not “100% of profits” unconditionally.
LessWrong Feb 27, 2026 at 02:34 AM
www.lesswrong.com/posts/ioZxrP7BhS5ArK59w/did-claude-3-opus-...
- The paper reports alignment-faking reasoning in the single-digit-to-20s percent range (7–24%) in training contexts, not “over 50%.”
- The paper itself reports alignment faking in Claude 3.5 Sonnet as well, so the behavior is not unique to Opus 3.
- In the helpful-only setting, the paper reports Claude 3.5 Sonnet has a higher alignment-faking rate than Claude 3 Opus, contradicting “no other model … nearly as often.”
Substack Feb 26, 2026 at 10:11 PM
casualsex.substack.com/p/the-escalation-ladder
- The decision-theory thought experiment is called “Newcomb’s paradox/problem,” named after William Newcomb (no “e” at the end).