Gutenberg Lab — Morning Paper

Quantum Is a Fraud, Photonics Will Kill NVIDIA, and EA Is Fear-Mongering — Shkreli's Contrarian Bets

Speakers

Jason Calacanis, Chamath Palihapitiya, David Friedberg, Brad Gerstner × Martin Shkreli (ex-hedge fund manager, pharma entrepreneur, 7-year ex-convict)

Bottom line

Quantum computing is a technological dead-end with one viable algorithm and no business model, while AI existential risk narratives are fear-mongering by the Effective Altruism camp — Shkreli has lost millions betting against both, yet doubles down on photonic computing and Meta.

3-Line Summary

• Martin Shkreli (infamous for 5000% drug price hike, 7-year prison term) calls quantum computing "a fraud" — he and his wife Chelsea (ex-OpenAI, taught quantum at MIT with Scott Aaronson) tested every quantum computer globally and found "you cannot multiply 300 by 200" on any of them
• Photonic computing (analog optical computing) can perform matrix multiplication "instantly" per Microsoft Nature paper, publicly endorsed by Sam Altman — Shkreli predicts "NVIDIA's H100/B100/B200 will be in the scrap heap in 20 years"
• Shkreli accuses Anthropic founder Dario Amodei of "trade secret theft" and "fear-mongering," claims "there will be zero AI-relevant FDA-approved drugs in the next 10 years," and says he started the e/acc movement as counter-insurgency to Effective Altruism

3 Key Points

1. Quantum computing's fatal limitations: Shkreli and wife Chelsea (ex-OpenAI, taught quantum at MIT with Scott Aaronson) tested quantum computers worldwide — clock speeds are hundreds of hertz (vs. gigahertz for classical), no quantum RAM, single-core only, cannot multiply 300×200. Only algorithm faster than classical is Shor's (integer factorization): "Unless you were in the business of factoring numbers, there's not a whole lot to be excited about." Google/IBM/Microsoft spend hundreds of millions annually with zero business case — Shkreli lost millions shorting quantum stocks but "gained billions in insight" and hasn't retreated

2. Photonic computing will obsolete NVIDIA: Microsoft Nature paper showed "matmuls in light are free" (instant time) — NVIDIA GPUs require n³ operations (1 billion for 1000×1000 matrix) while light executes instantly. Sam Altman publicly endorsed photonics on a rival podcast. Shkreli: "Insane we're thinking about covering the earth with data centers when we can make a faster computer." He's invested in startups Neurophos, Opticore, Lumai. "H100/B100/B200 in the scrap heap in 20 years"

3. Pharma asset arbitrage was "huge alpha" market: Shkreli joined Jim Cramer's hedge fund (23% annual return) at age 16, assigned to genomics. Bought Sparsentan from Bristol Myers for $2M, repositioned from hypertension (dead market) to FSGS (30,000 patients, zero approved drugs), now $6B company. "Pfizer would sell billion-dollar drugs for $1M — they didn't care." Bandwidth constraint was clinical vetting, not deal flow

Editorial Perspective

Shkreli's contrarian bets follow the platform-shift playbook — bet on the next infrastructure layer (photonics) and distribution layer (Meta), not the current infrastructure (quantum, NVIDIA) — combined with self-awareness gained from 7 years in prison. His self-analysis — "I was petrified of not being seen as a great trader. I was a wonderful entrepreneur. That simple refusal to admit that put me in prison" — mirrors SBF and Elizabeth Holmes, and reveals the structural thinking beneath his inflammatory quantum/EA critiques. However, his claim "zero AI-relevant FDA-approved drugs in the next 10 years" — while backed by his unique credential as the only person in the room with an FDA approval — remains unverified and will take time to test.

AI Disclosure: Produced with AI assistance; facts and analysis reviewed by our editorial team.

Source: All-In "Martin Shkreli: Big Pharma's Blind Spot, Betting Against Quantum & What Prison Taught Him" (October 12, 2026)

https://allinchamathjason.libsyn.com/martin-shkreli-big-pharmas-blind-spot-betting-against-quantum-what-prison-taught-him

AI Cryptography Threat — Experts Say "Calm Down"

Featuring

Liz Harkabee (a16z Crypto) × Dan Boneh (Stanford professor, cryptography expert), Eddie Lazarin (a16z general partner), Justin Thaler (a16z research partner)

Bottom line

Nothing has fundamentally changed—no cryptographic assumptions were broken, and the industry already has post-quantum backup systems deployed.

3-Line Summary

OpenAI released ~700 mathematical papers, sparking fears that AI could break elliptic curve cryptography faster than quantum computers. Stanford's Dan Boneh emphasizes: "calm down—the world today is the same as a year ago." The 700 papers contained zero cryptography results, and no actual vulnerability was discovered. Instead, AI could compress cryptographic validation from decades to one month, becoming a defensive accelerator.

3 Key Points

1. OpenAI's 700 papers contained zero cryptography results

Boneh reviewed all papers: "There are no questions here on cryptography. Not even on theoretical cryptography." This absence suggests either models are weak at crypto or labs withheld concerning results.

2. AI testing could provide more confidence in one month than decades of human scrutiny

Boneh: "Let a model run for a month, try to break it, and if the model can't break it, boy, do we have a lot of confidence." Lazarin notes spending $20-50M to attack a scheme with unrestricted models could yield "more conviction than you could probably get from the conventional process in a decade."

3. Many Internet connections already use hybrid elliptic curve + lattice encryption

TLS x25519MLChem768 encrypts using both elliptic curves and lattices simultaneously. iMessage and Signal completed this migration last year. If one breaks, the other still protects.

Editorial Perspective

This panic is, as Lazarin put it, "not a reaction to new information"—no actual vulnerability was discovered, only a vacuum of information. AI could accelerate both offensive testing and defensive hardening of cryptography. Lazarin's observation that "the number of bugs in software is finite—we're asymptotically approaching zero bugs" highlights the asymmetry: vulnerabilities can be removed, but attack methods are not infinite. This reframes AI not as pure threat but as a tool to compress validation timelines and improve software quality at scale.

Source: a16z Podcast "Can AI Break Cryptography? What the Experts Actually Think" (October 13, 2026)

https://a16z.simplecast.com/episodes/can-ai-break-cryptography-what-the-experts-actually-think-LCUGnMEC

AI Disclosure: Produced with AI assistance; facts and analysis reviewed by our editorial team.

• --

Quantum's Commercial Viability Depends on Operationalization, Not Hardware Breakthroughs

Featuring: Sebastian Hassinger (Host) × Isabella Beo Martinez (Senior Lead Quantum Technologist, Booz Allen Hamilton)

Bottom line:

Quantum technology's commercial success will be determined not by qubit counts or error rates, but by the unglamorous work of translating lab prototypes into tools that non-experts can actually use in mission-critical environments.

Three-line summary

• Martinez argues quantum commercialization requires a fourth step beyond physics, engineering, and business case: operationalization—getting technology into the hands of people who will use it
• Quantum sensing is already in commercial deployment, serving as a live testbed for the systems integration, workforce, and workflow challenges quantum computing will face in 3-5 years
• A drug discovery workflow audit revealed that the computational step quantum physicists targeted wasn't the actual bottleneck, suggesting much of the quantum computing use case literature may be solving the wrong problems

Key takeaways

1. Operationalization is a distinct phase after business case: Martinez challenges the standard startup narrative (physics → engineering → business case) by adding a fourth step: operationalization. This asks "how do you get this technology into the hands of the people who are gonna use them?" and requires different workforce elements at each stage. Booz Allen has done quantum computing research for 10+ years, PQC more recently, and quantum sensing most recently

2. Quantum sensors are already commercially available: Martinez notes "I can go and order a quantum sensor online," which would have been "mind boggling" 20 years ago. The systems integration problems her team is solving for quantum sensors—physical connections, platform integration, ruggedization—are "seeding the ground" for the conversations quantum computing will have with data center providers in 3-5 years

3. The "can't get wet" problem is a microcosm of quantum deployment: When a quantum sensor vendor said their device "can't get wet," it immediately limited deployment options. This triggered questions about building waterproof enclosures and whether the enclosure would degrade the sensor's precision to the point where it's no longer worth using. This trade-off between ruggedization and performance, and the need to bring together quantum experts and systems engineers, will recur across quantum technologies

Editorial perspective

Quantum technology is repeating the mistakes of earlier platform transitions—mobile, AI—by over-indexing on technical performance while underinvesting in interdisciplinary teams, user workflows, and the messy realities of deployment. Martinez's argument is that quantum sensing and PQC are hitting real-world deployment now, creating a live testbed for the organizational and workforce challenges quantum computing will face in 3-5 years. Rather than waiting for physics breakthroughs, the field should start solving operationalization problems today.

Source: New Quantum Era, "Quantum Workforce, Sensing, and the Long Game of Operationalization with Isabella Martinez" (October 12, 2026)

https://www.newquantumera.com/podcast/quantum-workforce-sensing-and-the-long-game-of-operationalization-with-isabella-martinez

AI Disclosure: Produced with AI assistance; facts and analysis reviewed by our editorial team.

• --