Every issue, newest first.
5 beats, 5 minutes, every weekday. Each story ends with the money play: the one move worth making.
Does Pasting Company Data Into ChatGPT Break Your NDA?
Pasting confidential data into ChatGPT can breach an NDA or expose a DTSA trade-secret claim depending on your NDA's wording and your account tier, because consumer ChatGPT trains on your chats by default while Team, Enterprise, and the API do not.
Is AI Training Fair Use? A 2026 Scoreboard
A 2026 scoreboard of AI-training copyright rulings: US courts split the fair use question on data provenance, holding training on lawfully obtained books fair use (Bartz v. Anthropic, Kadrey v. Meta) while pirated-library building was not (Anthropic's $1.5B settlement), with Ross v. Thomson Reuters, the UK Getty ruling, and Copyright Office guidance rounding out the picture.
Best Visa and Lowest Tax for a Remote AI Founder in 2026
For a profitable remote AI founder in 2026, the UAE gives the lowest personal tax (salary and personal investment income untaxed, company profit at 0% or 9%), ahead of Portugal's IFICI at 20% and Spain's Beckham Law at 24%, with residency day-count and where the company is run from ultimately deciding the bill.
Where to Incorporate an AI Startup: Delaware, Wyoming, or Estonia
A Delaware C-corp is the default for startups raising US venture money, Wyoming is cheaper with more privacy for bootstrappers, and an Estonian OU gives an online EU company, but none of these choices changes whether the EU AI Act applies, which follows where your AI output is used.
Can My Employer Force Me to Use AI at Work?
In an at-will US job your employer can generally require you to use AI, but the ADA, Title VII, and new California and Illinois rules let you demand accommodations, get notice, and challenge AI-driven employment decisions that discriminate.
How to design a protein with AI
A practical explainer on de novo protein design with the open-weight RFdiffusion, ProteinMPNN, and AlphaFold2 pipeline, runnable free in Google Colab, with the Nature paper's verified ~19% binder success rate (defined as binding at or above 50% of a positive control) and its roughly 100x gain over older Rosetta methods.
GNoME: Google DeepMind's 380,000 New Materials, and How to Pull Them Free
GNoME, Google DeepMind's deep learning model published in Nature in 2023, predicted 2.2 million crystal structures (381,000 newly stable), and you can pull roughly 380,000 of them free from Berkeley Lab's Materials Project or DeepMind's public Google Cloud bucket, with the data under a noncommercial CC BY-NC 4.0 license.
Which States Require an AI Hiring Bias Audit in 2026?
In 2026, New York City's Local Law 144 is the only US rule that clearly requires an independent annual AI hiring bias audit, while Colorado's audit-style law was suspended and rewritten (now effective 2027) and Illinois requires notice and non-discrimination rather than an audit.
When Will Fusion Power the Grid? The Real Dates Behind 2028 and 2035
No fusion plant powers a grid yet; the earliest contract-backed dates are Helion targeting 50 MW for Microsoft in 2028 and CFS's 400 MW ARC plant in Virginia in the early 2030s, while ITER will not run real fusion fuel until 2039.
Who Owns ChatGPT Output, and Can You Legally Sell It?
OpenAI's terms let you sell ChatGPT output and assign you whatever rights exist "if any," but in the US purely AI-generated work has no copyright, so real protection comes from adding and registering human authorship, not from the assignment itself.
Digital-Nomad Founder Taxes 2026: The FEIE, the 330-Day Rule, and the SE-Tax Trap
For 2026 the Foreign Earned Income Exclusion rises to $132,900, but it erases only federal income tax on foreign earnings, so a self-employed founder who passes the 330-day physical presence test still owes the full 15.3% self-employment tax unless a Totalization Agreement certificate of coverage or a properly structured entity applies.
EU AI Act Deadlines: The 4 Dates That Decide What You Owe in 2026-2027
The EU AI Act switches on across four dates (Feb 2 2025 bans, Aug 2 2025 general-purpose AI, Aug 2 2026 transparency plus high-risk, Aug 2 2027 product-embedded high-risk), and a 2026 Digital Omnibus proposal, still awaiting Official Journal publication, would push standalone high-risk to Dec 2 2027 and product-embedded to Aug 2 2028 while keeping the Article 50 transparency date fixed.
The AI frontier, in plain English. 5 minutes, every weekday.
One issue every weekday morning. Unsubscribe anytime.