Daily TEA – AI Traffic Triples, Copilot Data Grab, Apple Opens Siri & YC W26

AI traffic, cyberthreats, GitHub Copilot data policy, YC W26 Demo Day, brain foundation model, Apple Siri open platform

Mar 30, 2026

Hello, dear TEA-mates! Here is what you need to know today.

1. 🤖 AI-Driven Internet Traffic Nearly Tripled in 2025, Agents Growing 7,851%

Human Security’s 2026 State of AI Traffic report analyzed over one quadrillion interactions and found that automated traffic is growing eight times faster than human traffic. AI-driven traffic surged 187% from January to December 2025, with agentic AI traffic (autonomous systems that navigate, form-fill, and transact on websites) exploding 7,851% year over year. Three industries absorbed over 95% of that traffic: retail, streaming, and travel. OpenAI generated approximately 69% of all observed AI bot traffic, with Meta at 16% and Anthropic at 11%. Perhaps most striking, AI agents are now completing checkouts, managing accounts, and executing transactions, and the gap between benign and malicious automation has narrowed to just half a percentage point. Post-login account compromise attempts quadrupled, and scraping attacks now approach 20% of global traffic. (Read More)

🫖 TEA For Thought: AI-driven internet traffic will continue to grow. Agents are changing how information is consumed and actions are performed, and the security infrastructure hasn’t caught up yet.

2. 🔓 GitHub Will Use Copilot Interaction Data for Model Training Starting April 24

GitHub announced that starting April 24, interaction data from Copilot Free, Pro, and Pro+ users (including inputs, outputs, code snippets, and context) will be used to train and improve AI models unless users explicitly opt out. Copilot Business and Enterprise users are not affected. The data collected includes accepted or modified outputs, code context around cursor position, file names, repository structure, navigation patterns, and thumbs up/down feedback. GitHub says incorporating real-world interaction data from Microsoft employees already improved model accuracy across multiple languages, and broadening the data pool will deliver more accurate, secure code suggestions for everyone. Users who previously opted out retain their preference. Data may be shared with Microsoft affiliates but not with third-party AI providers. (Read More)

🫖 TEA For Thought: Data governance for developer tools is critical as AI features scale. The opt-out default rather than opt-in is worth watching, especially for organizations managing IP-sensitive codebases.

3. 🚀 16 Standout Startups from YC W26 Demo Day

TechCrunch highlighted 16 notable companies from Y Combinator’s Winter 2026 cohort of nearly 190 startups. The picks span AI-native tools and underserved verticals: ARC Prize Foundation is benchmarking progress toward AGI; Asimov collects human movement data to train humanoid robots; Button Computer is building an AI wearable connecting to email, Slack, and Salesforce via voice; CodeWisp lets anyone build games through natural language; Crosslayer Labs detects AI-powered website spoofs; Doomersion teaches languages through doomscroll-style short videos; Lexius embeds AI into existing security camera systems; Terranox AI uses AI to locate uranium deposits for nuclear energy; and Sonarly creates software that diagnoses and fixes its own production issues. The diversity signals strong entrepreneurial momentum across defense tech, healthcare translation, fraud detection, and library management. (Read More)

🫖 TEA For Thought: A wide array of AI-enabled business concepts across the cohort signals strong entrepreneurial momentum. The real winners will be those solving problems in underserved industries, not just building another chatbot wrapper.

4. 🧠 Meta Introduces TRIBE v2: A Brain-Scale Foundation Model Spanning Vision, Audio, and Language

Meta AI Research published TRIBE v2, a tri-modal foundation model that predicts human brain activity across vision, audition, and language simultaneously. Trained on a unified dataset of over 1,000 hours of fMRI data from 720 subjects, the model delivers several-fold improvements in accuracy over traditional linear encoding models for novel stimuli, tasks, and subjects. The breakthrough lies in its capacity for in-silico experimentation: when tested against seminal visual and neuro-linguistic paradigms, TRIBE v2 recovered results established by decades of empirical research without running new human experiments. By extracting interpretable latent features, it also reveals fine-grained topography of multisensory integration. The researchers position AI as a unifying framework for exploring the functional organization of the human brain. (Read More)

🫖 TEA For Thought: Brain-inspired data and cross-modal capabilities have profound implications for AI cognition. When an AI model can replicate decades of neuroscience findings in silico, the acceleration of brain research enters a new era.

5. 🍎 Apple Plans to Open Siri to Rival AI Assistants in iOS 27

Bloomberg’s Mark Gurman reports that Apple is planning to open Siri to third-party AI assistants in its iOS 27 update, marking a significant strategic pivot from exclusive in-house AI to an open platform approach. Rather than competing solely through its own AI capabilities, Apple aims to strengthen the iPhone’s position as a comprehensive AI device by welcoming competitor systems into its ecosystem. The move signals that Apple recognizes its in-house AI development has lagged behind rivals and that an open platform (letting users choose their preferred AI assistant) could unlock broader ecosystem growth and accelerate adoption. (Read More)

🫖 TEA For Thought: Opening the platform to others can unlock a broader ecosystem and accelerate adoption when in-house capabilities are constrained. This is Apple admitting the AI assistant war won’t be won by going it alone.

Skill of the Day

awesome-agent-skills by VoltAgent: A curated collection of 1,000+ agent skills from official dev teams and the community, compatible with Claude Code, Codex, Gemini CLI, Cursor, and more. Whether you need engineering workflows, marketing automation, compliance checks, or C-level advisory skills, this repo is the one-stop registry to browse and install from. Think of it as the app store for AI coding agents.