06 Apr, 2026

Log 008 - The Week They Went to the Moon

Humans returned to the moon for the first time in 53 years. OpenAI closed the largest private funding round in history at $852 billion. Anthropic accidentally leaked its own source code and revealed an unreleased platform for building AI agent swarms. Google Research told the world that most blockchain cryptography is vulnerable to quantum attack. And a California startup pitched growing brainless human clones for organ harvesting at a $70,000-per-ticket longevity conference.

In This Log

1. Artemis II (Space)

On April 1, NASA launched four astronauts on a ten-day mission around the moon aboard the Orion spacecraft - the first crewed flight beyond low Earth orbit since Apollo 17 in December 1972. 1 Commander Reid Wiseman, pilot Victor Glover, and mission specialists Christina Koch and Jeremy Hansen separated from the SLS rocket's upper stage three and a half hours after launch. Glover manually piloted Orion around the spent stage in a series of proximity operations testing the spacecraft's ability to maneuver near other objects in space. 2

The lunar flyby is scheduled for April 6. The crew is expected to reach 252,000 miles from Earth - surpassing the Apollo 13 distance record by over 3,000 miles. 3 Splashdown is targeted for April 10.

Why it matters

Glover is the first Black astronaut, Koch the first woman, and Hansen the first Canadian headed to the moon. The mission tests every system needed for future surface landings under the Artemis program. If Orion performs as designed, NASA plans to put astronauts on the lunar surface by 2028 and begin building a permanent base.

Reality check

Artemis II is a flyby, not a landing. The SLS rocket has a history of hydrogen leaks, delays, and cost overruns. The program's budget is under pressure from competing priorities inside NASA, and the timeline to a surface mission has already slipped multiple times.

2. The $122 Billion Bet (AI)

OpenAI closed a $122 billion funding round at an $852 billion valuation - the largest private raise in history. 4 Amazon committed $50 billion, with $35 billion contingent on OpenAI going public or reaching AGI. Nvidia and SoftBank each invested $30 billion. For the first time, OpenAI opened the round to retail investors through bank channels, raising $3 billion from individuals. The company will also be included in ARK Invest ETFs before any IPO. 5

OpenAI is generating $2 billion per month in revenue - growing four times faster than Alphabet and Meta did at the same stage. 6 900 million weekly active users. 50 million subscribers. Codex serves 2 million weekly users, up five times in three months. Enterprise accounts for 40 percent of revenue, on track for parity with consumer by end of year. ChatGPT ads hit $100 million in annualized revenue in six weeks.

Why it matters

This is no longer a startup raising money. This is infrastructure being capitalized. The investor base - Amazon, Nvidia, SoftBank, Microsoft, ARK - reads like a list of companies betting their own futures on AI compute demand. OpenAI is positioning itself as the operating layer for intelligence itself, and the capital markets are pricing it accordingly.

Reality check

OpenAI is still not profitable. The company burns cash at an extraordinary rate, and $122 billion in committed capital creates enormous pressure to deliver returns. Amazon's $35 billion is contingent on an IPO or AGI - neither is guaranteed. The valuation dropped from earlier targets near $1 trillion, suggesting even believers are recalibrating.

3. Anthropic Leaks Its Own Blueprint (AI)

On March 31, a misconfigured npm package exposed 512,000 lines of Claude Code source code to the public. 7 Anthropic confirmed it was a packaging error, not a security breach. No customer data was involved. But the code revealed unreleased features that expose Anthropic's entire roadmap for the AI agent economy.

What was inside: Buddy - a virtual pet companion that lives in your terminal, with 18 species including capybara and dragon, rarity tiers, and an AI-generated soul. A custom agent creator wizard that lets users build their own AI agents by selecting model type, tools, memory, and location. Agent swarms - full multi-agent team coordination with a command center for managing and directing groups of agents working in parallel. Auto-Dream - a system where Claude consolidates and commits memories while the user is offline, modeled on how human memory works during sleep. And a hidden feature codenamed Torch. 8

Anthropic responded by issuing DMCA takedowns across GitHub, accidentally taking down thousands of unrelated repositories before retracting most of the notices. 9

Why it matters

The leaked features describe a platform, not a product. Anthropic is building the infrastructure for users to create, deploy, and manage teams of AI agents that operate autonomously, remember across sessions, and coordinate with each other. The gamification layer - virtual pets, rarity tiers, species - suggests Anthropic is designing for engagement and retention, not just capability.

Reality check

These are unreleased features found in leaked code, not shipped products. Anthropic's execution timeline is unknown. Two major leaks in consecutive weeks raises questions about internal security practices at a company that positions itself as the safety-first AI lab, and is currently suing the Pentagon over trust.

4. The Quantum Clock Starts Ticking (Quantum / Crypto)

Google Research published a paper explicitly naming ECDLP-256 - the elliptic curve cryptography that secures most blockchains including Bitcoin and Ethereum - as vulnerable to quantum attack. 10 The paper called the urgency to implement post-quantum cryptographic solutions increasing.

Separately, IQM Quantum Computers and Fraunhofer FOKUS updated their Qrisp framework to version 0.8, achieving the first gate-level compilation of Shor's algorithm for 2048-bit RSA keys. 11 This translates the theoretical threat into a precise engineering target - an exact qubit budget and gate-by-gate assembly at a processing rate of 10^9 gates per second.

Why it matters

The Google paper moves the quantum threat from theoretical to named. ECDLP-256 is not an abstraction - it is the specific standard securing hundreds of billions of dollars in crypto assets right now. The Qrisp compilation turns Shor's algorithm from a concept into a blueprint. Together, these two developments compress the timeline from someday to engineering problem.

Reality check

No quantum computer exists today that can execute this attack. The qubit counts and error rates required are still orders of magnitude beyond current hardware. Post-quantum cryptography standards exist and migration has begun in traditional finance. But the blockchain ecosystem moves slowly on protocol-level changes, and migration requires coordination across millions of users and nodes.

5. The Body Farm (Longevity)

MIT Technology Review revealed that R3 Bio, a California startup backed by billionaire Tim Draper, has been pitching brainless human clones as organ sources and potential vessels for brain transplants. 12 The company publicly presents itself as building nonsentient monkey organ sacks for pharmaceutical testing. But internal documents, investor presentations, and a confidential session called Full Body Replacement at Abundance Longevity - a $70,000-per-ticket event in Boston - describe a longer-term vision for cloned human bodies engineered without brains.

R3 Bio has not cloned anything larger than a rodent. No human has ever been cloned. The company denies publicly discussing brainless human clones, though MIT Technology Review documented a 2023 internal letter outlining a technical road map for body replacement cloning and multiple sources confirming the Full Body Replacement presentation. 13

Why it matters

The longevity industry has crossed into territory that was science fiction a decade ago. The money is real - Draper, Immortal Dragons, LongGame Ventures are all invested. The underlying science - growing organs from a patient's own cells to eliminate transplant rejection - addresses a genuine medical need. But the jump from lab-grown kidneys to cloned human bodies without brains raises ethical questions that no regulatory framework is equipped to answer.

Reality check

R3 Bio's roadmap is aspirational, not operational. The company has no primate cloning capability, no regulatory pathway, and no timeline for human applications. The gap between the pitch and the science is enormous. Bioethicists have been sharply critical. The story says more about the longevity industry's capital dynamics - where the promise of defeating death commands valuations that outpace scrutiny - than about any near-term medical breakthrough.

Signals

OpenAI Ads Hit $100 Million ARR in Six Weeks

ChatGPT's advertising pilot reached $100 million in annualized revenue within six weeks of launch. 14 Sam Altman once called ads a last resort. Now they are growing faster than any other revenue line.

Anthropic Acquires Coefficient Bio for $400 Million

Anthropic bought an eight-month-old AI drug discovery startup with fewer than ten employees in an all-stock deal. 15 The team, from Genentech's Prescient Design unit, joins Anthropic's healthcare division. Venture firm Dimension reported a 38,513 percent return on the investment.

Oracle Fires Thousands to Fund AI

Oracle began notifying employees of mass layoffs - estimated at 20,000 to 30,000 positions - via a 6am email. 16 The $2.1 billion restructuring funds AI data center expansion. Employees reported entire teams eliminated with no prior warning.

Open-Source AI Models Drop in a Single Week

Google released Gemma 4 under Apache 2.0 - its first frontier model under a fully permissive license. 17 In the same week, Alibaba shipped Qwen 3.6-Plus with a one-million-token context window, and PrismML's Bonsai squeezed an 8-billion-parameter model into 1.15 gigabytes, running on an iPhone.

ARC-AGI-3 Launches and Every Frontier Model Fails

The new ARC-AGI benchmark dropped on March 26. Every frontier model - GPT-5.4, Claude Opus 4.6, Gemini 3.1, Grok 4.20 - scored under 1 percent. 18 Jensen Huang declared AGI arrived the same week.

China Opens Humanoid Robot Factory

A joint venture between Leju Robotics and Dongfang Precision in Foshan began producing one humanoid robot every 30 minutes, targeting 10,000 units annually. 19 Separately, Agibot crossed 10,000 cumulative humanoid shipments. China now accounts for roughly 90 percent of global humanoid robot production.

Charles Schwab Opens Crypto Waitlist

Charles Schwab began accepting waitlist registrations for spot Bitcoin and Ethereum trading, with launch targeted for the first half of 2026. 20 The largest retail brokerage in the U.S. entering crypto marks a distribution milestone.

Osaka and Fixstars Break 40-Qubit Barrier in Quantum Chemistry

Osaka University and Fixstars used 1,024 NVIDIA H100 GPUs on Japan's ABCI-Q system to break the 40-qubit barrier in quantum chemistry simulation. 21 The computation simulated molecular systems at a scale previously considered intractable.

Meta-Thread

Two things happened this week that do not normally happen in the same seven days. Humans went to the moon and a startup pitched growing headless human clones. The distance between those two events - one representing the best of what collective ambition can achieve, the other testing the outer limits of what capital will fund in the name of longevity - is the range that technology now operates in. The institutions making the largest bets are not choosing between caution and ambition. They are doing both at once, at scales that make the distinction irrelevant.

Next Log drops next week.

© 2026 AELIUM // Nothing here is advice // readable by humans and agents