In Partnership With
October 20, 2026 · READ ONLINE
Keeping you ahead in AI.
THE DAILY BRIEFING / 4 MIN READ
Welcome back, AI enthusiasts.
Elon Musk wants every American child to own a small piece of SpaceX. He’s promised one share each next year, a gift estimated at nearly $11 billion at recent prices, shortly after Anthropic announced its own public benefit fund. Let’s Dive In!
IN TODAY'S ISSUE
01 Elon Musk Pledges Own SpaceX Shares to America’s Children
02 Amazon Takes Aim at Nvidia’s Biggest Software Advantage
03 GPT-6 Gets an Engineering Rival That’s 150× Cheaper
01 / PUBLIC STOCKS
Evolving AI: Elon Musk plans to give every child in America one of his own SpaceX shares in 2027 and says more gifts will follow.
KEY POINTS
At roughly $146 per share, a gift covering about 73 million children would be worth nearly $11 billion, depending on the eventual share price and eligibility.
President Trump thanked Musk for the planned gift through Trump Accounts, the government’s tax-advantaged investment accounts for children.
Musk has not published eligibility rules or an exact delivery date, and has committed no fixed dollar amount or percentage of SpaceX.
DETAILS
Elon Musk’s plan would make every American child a shareholder in SpaceX, which also owns the AI company xAI. New Treasury rules allow donated stock in Trump Accounts, generally requiring shares to be held for five years or until the start of the year the child turns 18, whichever comes first. Musk recommended giving shares directly to the public after Anthropic proposed committing 5% of its shares to a public benefit fund, subject to shareholder approval.
WHY IT MATTERS
Musk’s pledge could raise expectations for other AI leaders to share more of the wealth their companies create. If AI pushes company values higher while making some jobs less secure, owning a stake could become more important to families’ financial futures. Giving shares offers a practical way for more people to benefit from that growth. The lasting value will depend on how much ownership reaches them and whether contributions continue, since families would need more dependable support than occasional gifts to offset any sustained loss of earnings.
TOGETHER WITH GUIDELESS
Evolving AI: As companies grow, so does the amount of software people need to learn: internal tools, customer platforms, workflows, processes, and everything in between.
The usual solution is more documentation, more videos, more onboarding calls, and more people answering the same questions. Guideless gives teams a more scalable way to do it.
Capture a workflow once, and Guideless turns it into polished, reusable training content that can be updated, shared, embedded, and managed from one place.
That means teams can:
Create software guides up to 12x faster
Let AI handle up to 90% of the creation work
Produce narrated training in 40+ languages
Keep guides updated without starting from scratch
Give employees and customers more ways to help themselves
Guideless is already being used by organizations across employee training, L&D, Customer Success, and Support, and the bigger vision goes beyond just creating guides. By turning captured workflows into reusable operational knowledge, Guideless is building toward a world where that same knowledge can power different training formats, improve processes, and eventually help both people and AI agents understand how work gets done.
Less time documenting. Less time explaining. More knowledge that scales.
02 / AI HARDWARE WAR
Source: Amazon News
Evolving AI: Amazon has introduced TorchNeuron, a new tool that makes its own AI chips easier to use, challenging Nvidia's long-standing software advantage.
KEY POINTS
Now in public beta, TorchNeuron lets developers run existing PyTorch AI models on Amazon's Trainium chips with minimal code changes.
The software supports 89 of Hugging Face's 100 most-used text-generation model architectures, up from 46 with Amazon's previous integration.
Voice AI company Deepgram says Trainium processed about 60% more audio per dollar than its existing GPUs in its benchmarks.
DETAILS
Amazon's new tool targets Nvidia's CUDA software ecosystem, which has become a major reason AI developers stick with its GPUs. AWS began renting Nvidia chips in 2010, helping more developers access powerful computing without buying the hardware. TorchNeuron now uses PyTorch's built-in hardware support to bring familiar training and debugging tools to Amazon's chips. The software is still in beta, while AWS continues offering Nvidia GPUs alongside its own alternatives.
WHY IT MATTERS
Amazon helped bring Nvidia's GPUs within reach of more developers through AWS, contributing to their wider adoption as AI took off. Now, it has a strong financial incentive to move more AI workloads onto its own chips, reducing reliance on outside suppliers and gaining greater control over costs. Google is pursuing a similar approach, giving AI companies more hardware options to experiment with. As these alternatives become easier to use, competition could intensify over pricing, performance, and availability, giving AI labs more negotiating power. If that competition brings down the cost of training and running AI models, it could eventually lead to cheaper AI services, although those savings are not guaranteed to reach users.
03 / SPECIALISED AI
Evolving AI: Skyfall has introduced SKY-001 to tackle the engineering work behind complex objects such as cars and engines.
KEY POINTS
In Skyfall’s computer-aided design tests, SKY-001 generated a component in 3.9 seconds for $0.005 versus GPT-6 Astra’s 199 seconds and $0.75.
Across 1,000 components, it scored 79.7% against Astra’s 79.1% on shape overlap, which measures how closely the generated geometry matches a reference.
SKY-001 also generates building floor plans with rooms, doors, and windows, taking about 1.2 seconds per plan in Skyfall’s tests.
DETAILS
Skyfall is trying to automate the engineering work that turns an idea into a physical design. Its demos compare SKY-001 with GPT-6 Astra, Claude Opus 5.5 and Claude Fable 5.1, which were asked to generate whole assemblies in a single request. SKY-001 plans individual parts from a reference image, including internal components hidden from view, then predicts the sketches and 3D steps needed to shape each piece. Separate parts can be processed in parallel. Editable CAD output remains a next step for the research preview, with commercial deployment targeted for early 2027.
WHY IT MATTERS
Skyfall’s results suggest specialist AI companies could compete with the biggest labs by making specific professional tasks much cheaper. If those gains hold in everyday engineering work, teams could explore more designs within the same budget before choosing which ones to prototype. That gives companies a practical reason to consider models built for their profession. The savings will depend on whether engineers can edit and validate the output without spending the time they saved correcting it.
04 / HUMAN OR MACHINE?
One is real. Trust your eyes?
Two images. One photograph. Click the one you think is real.
Image 1
Image 2
05 / QUICK HITS
🧬 Harvard-affiliated researchers developed an AI facial analysis model trained on 40M+ images to estimate biological age and cancer survival risk.
🇨🇳 China proposed new export controls on AI, including large models and advanced AI technologies, with temporary registration rules already in effect.
⚡ NVIDIA's Vera CPU doubled AI-agent task throughput in Modal's early tests, outperforming competing server processors under heavy load.
🧠 MiniMax launched its M3.1 Flash model, claiming nearly 3× faster responses and 70% lower costs for coding and agent tasks.
06 / TRENDING AI TOOLS
🎞️ Guideless - AI tool that automatically turns your clicks and workflows into step-by-step video guides*.
📣 MentionBird - Shows why AI recommends competitors and what sources influence answers.
🎬 Rushilo - Edit videos by telling Claude Code or Codex what to change.
💻 Aperture - Open-source AI code editor that checks changes before applying them.
*PARTNER LINKS
BEFORE YOU GO
Worth your four minutes?
Help us make the next edition better.
See you tomorrow,
The Evolving AI team
