Byte

Real-time reporting on AI, startups, defense tech, space, and the forces shaping technology. Sign up:

[Byte @ByteEcosystem]

NEW - 🇻🇦 Rubio arrives in Vatican for talks with pope: AFP

[Byte @ByteEcosystem]

Anthropic launched 10 AI agent templates for finance teams, covering KYC checks, monthly close reports, and presentation prep. The templates are available in Claude Code and Cowork on paid plans, with new Excel, PowerPoint, Word, and Outlook integrations.

[Byte @ByteEcosystem]

xAI will merge into SpaceX and be rebranded as SpaceXAI, with xAI projects folded into SpaceX products. No further restructuring details have been provided.

[Byte @ByteEcosystem]

Coinbase is testing 1-person product teams supported by AI agents, with one operator handling coding, design, and parts of product work. The move follows a roughly 14% workforce cut and a flatter org under CEO Brian Armstrong.

[Byte @ByteEcosystem]

Anthropic says it partnered with SpaceX for additional compute, doubled Claude Code usage limits across plans, ended peak-hours reductions, and raised API limits for Opus.

[Byte @ByteEcosystem]

Panthalassa, founded by former Apple, SpaceX, NASA, and Tesla engineers, unveiled floating data centers powered by wave energy. The offshore platforms would use satellites to support AI services, with commercial launch planned for 2027.

[Byte @ByteEcosystem]

NEW: Ted Turner, CNN founder, dead at 87.

[Byte @ByteEcosystem]

ChatGPT is now available as an add-on in Excel and Google Sheets, with support for data analysis, formula writing, spreadsheet updates, and step-by-step explanations. Powered by GPT-5.5.

[Byte @ByteEcosystem]

OpenAI is rolling out GPT Instant 5.5 to all ChatGPT users, with shorter responses, fewer hallucinations, stronger reasoning, and a clearer memory interface.

[Byte @ByteEcosystem]

NEW - Brent crude oil falls over 11% to $101.

[Byte @ByteEcosystem]

Study: A 2026 Dermatology and Therapy trial of 60 people found 8 weeks of topical DHM reduced biological skin age and wrinkles, potentially reversing skin aging epigenetically via DNMT1 inhibition and altered DNA methylation.

[Byte @ByteEcosystem]

Boston Dynamics shared new footage of its production electric Atlas robot rotating its body while balancing on its arms, highlighting movement control and a hand adjustment for stability.

[Byte @ByteEcosystem]

NEW - 🇺🇸 Average US gas price rises above $4.50 per gallon for the first time since 2022.

[Byte @ByteEcosystem]

Uni-1.1 image gen API from @LumaLabsAI is here.

This unlocks new experiences you can build when visual reasoning becomes programmable.

One app idea: time-travel GeoGuessr.

Enter a scenario like “New York in 1910” or “a Martian colony in 2180.”

Uni-1.1 generates a panoramic image, and our app turns it into an explorable 3D scene, and you guess where + when you are. Really cool use case.

Start shipping → https://t.co/dUHIi5pHae

#LumaPartner

[Byte @ByteEcosystem]

Sam Altman says "doom sells" because people pay more attention to bad news, and argues AI CEOs warning that 50% of jobs will disappear while building companies are being tone deaf.

[Byte @ByteEcosystem]

Google, Microsoft, and xAI will give the U.S. Commerce Department early access to unreleased AI models so government experts can assess capabilities, security, and risks before public release.

[Byte @ByteEcosystem]

@alex_whedon Awesome!

[Byte @ByteEcosystem]

SubQ is launching the first production-ready sub-quadratic LLM: 12M-token context, 52x faster than FlashAttention, and better long-context performance than Opus 4.6 at under 10% of the cost.

[Alexander Whedon @alex_whedon]

Introducing SubQ - a major breakthrough in LLM intelligence.

It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA),

And the first frontier model with a 12 million token context window which is:

- 52x faster than FlashAttention at 1MM tokens
- Less than 5% the cost of Opus

Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention).

Only a small fraction actually matter.

@subquadratic finds and focuses only on the ones that do.

That's nearly 1,000x less compute and a new way for LLMs to scale.

[Byte @ByteEcosystem]

Anthropic reportedly withheld Mythos over safety concerns, but a small private Discord group allegedly accessed it on day one by guessing a URL tied to naming conventions exposed in the Mercor breach and using a contractor’s eval credentials.

[Byte @ByteEcosystem]

NEW - 🇺🇸 U.S. government has ~$26.5B unrealized gain after 9.9% Intel stake jumps 22% following Q1 results.