04-19-2026Price:

The Frontier

Your signal. Your price.

SCIENCE

NIH whistleblower exposes Alzheimer's data fraud

Sunday, April 19, 2026 · from 3 podcasts
  • A single manipulated image in a 2006 Nature paper shaped 30 years of failed Alzheimer’s research.
  • NIH officials ignored fraud while directing $2.7B toward a discredited theory.
  • Whistleblower reveals waste clearance, not amyloid, may be the real key to treatment.

The foundation of Alzheimer’s research cracked in 2022 when neuroscientist Matthew Shrag uncovered doctored Western blots in a landmark 2006 Nature paper by Sylvain Lesné. The image manipulation wasn’t minor - it fabricated evidence for a specific amyloid oligomer said to cause cognitive decline. That paper became a cornerstone of the amyloid cascade hypothesis, which dominated research for decades.

The fraud extended far beyond one study. Investigative journalist Charles Pillar found over 100 papers by NIH official Eliezer Masliah contained similar image irregularities - some stretching back 30 years. Masliah oversaw a $2.7 billion federal budget. When the data is manufactured, the science is over before it begins.

For three decades, the amyloid hypothesis funneled billions into drugs like Aduhelm. They cleared plaques but failed patients. Pillar calls it a hijacking - a monoculture of researchers protecting a failing dogma. Careers, grants, and prestige all depended on amyloid. Alternatives were starved of funding.

Shrag argues the real problem isn’t plaques - it’s the brain’s failure to clear waste. He points to blood vessel health and blood pressure control as more promising paths. Yet NIH and universities suppressed dissent. The University of Minnesota delayed investigating Lesné for years, letting him keep federal funding.

"The most important thing is not to fool yourself - and you are the easiest person to fool."

- Matthew Shrag, Freakonomics Radio

Regulatory capture deepened the crisis. Pillar exposed FDA officials who approved amyloid drugs then took board seats at the same companies. Science became a pipeline for profit, not truth. Whistleblowers were isolated. Shrag found his own mentor, Othman Ghribi, had manipulated images in their joint work - labeled as 'exaggeration' to make results clearer.

"We’ve been chasing a phantom for 30 years while real patients deteriorated."

- Charles Pillar, Freakonomics Radio

The cost is measured in wasted lives and misdirected funding. The amyloid theory didn’t just fail - it actively blocked progress. Now, with the fraud exposed, the field must confront a simple question: how much longer will institutions protect the past instead of pursuing real cures?

Source Intelligence

- Deep dive into what was said in the episodes

671. Why Has There Been So Little Progress on Alzheimer’s Disease?Apr 17

  • Charles Piller argues the amyloid hypothesis has dominated Alzheimer's research since 1990, directing tens of billions in funding toward drugs that remove beta-amyloid plaques but fail to arrest cognitive decline.
  • Matthew Schrag states anti-amyloid antibody drugs like aducanumab are dangerous, causing brain swelling and bleeding, and offer only imperceptibly subtle cognitive benefits. He notes aducanumab was withdrawn for being ineffective and dangerous.
  • Piller and Schrag's investigation found apparent image manipulation in 132 of 800 papers by influential NIH neuroscientist Eliezer Maslia, tracing problems back 30 years. The NIH made no comment when Maslia left his post.
  • A seminal 2006 Nature paper by Sylvain Lesné and Karen Ash, which proposed a specific amyloid oligomer as the toxic cause of Alzheimer's, was retracted after Schrag and Piller found its Western blot images were severely manipulated to support the hypothesis.
  • Schrag discovered his mentor, Othman Ghribi, had manipulated images in their joint research, describing it as 'exaggeration' to make results clearer. Multiple papers were retracted, and Ghribi stated he took full responsibility as lab director.
  • Piller cites a Public Citizen report concluding regulatory capture has infiltrated the FDA, noting 11 of 16 FDA examiners for Alzheimer's drug approvals left to work for the companies they regulated.
  • The NIH spends about $4 billion annually on Alzheimer's and dementia research, second only to cancer spending and up from $1 billion a decade ago.
  • Schrag reformulates Alzheimer's as a disease of failed waste clearance in the brain, arguing a broader approach targeting blood vessel health and aggressive blood pressure control shows more promise than singular amyloid focus.
  • Casava Sciences paid a $40 million SEC settlement for misleading investors about its drug simufilam, which failed clinical trials. Scientist Hoau-Yan Wang was indicted for data fabrication but charges were later dropped.
Also from this episode: (1)

Health (1)

  • Alzheimer's affects over 7 million people in the U.S., with higher prevalence and earlier onset linked to pollution exposure, lower educational attainment, and economic inequality.

Nostr Compass #17Apr 16

Also from this episode: (25)

AI & Tech (11)

  • Amethyst transitioned to RT Tor, a Rust implementation of the Tor network, resolving previous crashing issues caused by outdated cTor bindings.
  • Amethyst version 1.7.3 improved video playback UI and migrated badges and bookmarks to new NIP kind numbers for compatibility with updated standards.
  • Amethyst is re-implementing MLS and Marmot protocols and a new secp256k1 library in pure Kotlin, based on published test vectors, aiming for encrypted direct messaging and Schnorr signature support.
  • Amethyst is developing NIP-AC for peer-to-peer WebRTC voice/video calls over Nostr, with two competing approaches being WebRTC and the newer, potentially more performant media over QUIC.
  • Nostralgia version 1.27 introduced video recording and editing, animated GIF support for profiles, new keyboard shortcuts, and private replies within direct messages.
  • Sandré is developing Nostria version 4 with integrated local AI models, including Google's Gemma 4 running in-browser for chat and image generation, and future plans for Nostr-based AI agents.
  • Titan version 0.1.0 launched a new internet browser with native Nsite capability, allowing pubkeys to publish websites via Nostr events and registering names permanently on-chain using Bitcoin transaction op-returns.
  • Bickl version 1.5.0, a decentralized cycling tracker, added background GPS tracking for GrapheneOS and integrated eCash transactions via Nutsapps, enabling scavenger hunt features with digital tokens.
  • Sprout, a Slack-like communications platform by Bloq, uses Nostr and Blossom for chat rooms, NIP-42 for authentication, NIP-29 for group management, and natively integrates AI agents via a Goose instance.
  • Mesh LLM version 0.56.0, a distributed LLM inference system, uses Nostr key pairs for node identities and coordinates AI computation across machines, presenting potential for L402 Lightning payments.
  • NostrVPN added exit node support, allowing servers to act as privacy-enhancing exit nodes for WireGuard tunnels, synchronizing invites and aliases over Nostr, and released an Umbrel package.

Social Media (1)

  • Social version 0.15 allows scheduling recording sessions for live streams and enables viewers to create TikTok-style vertical video clips from shows.

AI Infrastructure (2)

  • Note-Tek version 0.10 Beta integrated the Zap Store's Nostr-based app update mechanism, enabling automatic application updates, which is generally good for security but raises concerns about user control and enshittification.
  • White Noise switched from BlurHashes to ThumbHashes for image previews, significantly reducing size while increasing quality, and progressed on NIP55 push notifications and cursor-based chat message pagination.

Open Source (3)

  • Amber's 6.00 pre-release introduced per-connection keys for NIP-46 remote signing, enhancing security by isolating compromises, and also added automatic updates via the Zap Store NIP.
  • Nostria released a new native mobile app using the Tauri Rust framework for Linux, macOS, and Windows, featuring improved Amber/Aegis NIP55 signer integration and a focus on social sharing previews.
  • OpenSats announced its 16th wave of Nostrgrants, funding Amethyst Desktop, Nostermail, Nostrord (a Kotlin multiplatform NIP-29 client), Nuru Nuru (a Japanese client), and renewing Hamster's grant.

Protocol (8)

  • Nymchat reverted its Marmot protocol integration for NIP-17 group chats due to the lack of proper multi-device support, highlighting the importance of this feature for cross-platform clients.
  • NUC, the Nostr Army Knife, updated to version 0.19.5 with Blossom multi-server support for parallel uploads/downloads, a new key command, and native support for the outbox model.
  • Snort's security audit led to fixes for Schnorr signature verification and NIP46 relay message forgery protection, alongside improvements in PIN encryption, batch verification, and lazy loading.
  • Relator, a Web of Trust scoring engine, introduced a scheme for writing validators in the ELO functional programming language, publishing them as Kind 765 events on Nostr, and launched a plugin marketplace.
  • A proposal for NIP 24 suggests adding `published_at` and `created_at` timestamps to all replaceable events, distinguishing original creation from last edit, which would simplify tracking user join dates.
  • NIP-17 for private direct messaging replaced NIP4 to address metadata leakage and weaker encryption, utilizing NIP44 for encryption and NIP59 (Giftwrap) for metadata protection, with randomized `created_at` timestamps.
  • NIP-17's Giftwrap uses an unsigned 'rumor event' (Kind 14) wrapped in a signed 'seal' (Kind 13) and then an outer ephemeral 'seal' (Kind 1059) to protect sender identity, though it lacks post-compromise or forward secrecy.
  • NIP-46, the Nostr remote signing protocol (NSEC-Bunker), allows clients to request signatures from an application holding the NSEC via Kind 24133 events and NIP44 encryption, using JSON-RPC-like structures.

Technology, Culture, and the Next AI Interface with signüllApr 16

  • Signüll criticizes regulatory moves like New York State potentially banning AI for health or financial advice, arguing it sets back average consumers while leaving the wealthy with human advisors unaffected.
Also from this episode: (11)

AI & Tech (10)

  • Signüll argues technology's current acceleration feels like hitting a 100x speed button on a simulation, compressing events that happened a month ago to feel like a decade past.
  • Signüll sees the primary challenge for AI is making model power accessible and useful for individuals, a shift he believes is beginning with agents but remains primitive and inaccessible.
  • Signüll frames his online commentary as exploring the intersection of technology and culture, relating life and culture back to computer science principles he learned as a kid.
  • Signüll believes the current technology cycle is uniquely hard because it involves developing AI personality, which he calls an insane shift from building delivery vehicles for human content.
  • Anish Charya notes a study showing AI is highly popular in China but has a negative Net Promoter Score in the U.S., where it is less popular than ICE vehicles.
  • Charya argues the way to fix AI's negative NPS is to make important things like healthcare and education cheaper quickly using AI, not just reducing inflation but causing actual deflation.
  • Charya claims 45% of healthcare costs are administrative overhead, and education costs could be cut by restoring student-administrator ratios to levels from ten years ago and making professors modestly more productive.
  • Charya contrasts healthcare and education as intelligence-bound problems solvable by AI with housing, which he calls a collective action problem unrelated to technology.
  • Signüll observes that most people use AI for very basic tasks despite advanced capability demonstrations, and the industry is still in the stone ages of user perception and utility.
  • Signüll suggests giving normal people equity stakes in major AI companies like OpenAI and Anthropic could create an ownership mentality and improve public sentiment toward AI.

Startups (1)

  • Signüll teases he is building a small, fun consumer AI product focused on creating an interesting, out-of-the-box experience for average people.