Get tomorrow's brief in your inbox
Today: Google released Gemini 3.1 Pro with benchmark scores rivaling Claude Opus 4.6 at half the price. Accenture told senior staff to use AI tools or lose out on leadership promotions. India pulled in $200B+ in AI infrastructure commitments at this week's summit.
Google's Gemini 3.1 Pro is now available in preview, and the pricing could reshape how small businesses think about AI spending. At $2 per million input tokens and $12 per million output tokens, it's less than half the price of Claude Opus 4.6 while delivering similar benchmark scores. For businesses running heavy AI workloads, automated customer support, content generation, or data analysis workflows, this means you can potentially cut your AI infrastructure costs in half without sacrificing capability.
The model is showing strong performance on real professional tasks. AI startup Mercor's benchmarking system (which tests how well AI handles actual knowledge work) ranked Gemini 3.1 Pro at the top of their leaderboard. If you've been putting off AI adoption because of cost, or if you're already using premium models and watching your bills climb, this is worth testing. The catch: launch day demand is causing slowdowns and errors, so wait a few days before running critical workflows through it. Google Gemini 3.1 Pro release
Plaud now integrates with Zapier for automated workflow routing. Plaud is a wearable AI notetaker that records and transcribes in-person conversations, not just Zoom calls. The Zapier integration means transcripts can auto-archive to Google Drive, push summaries to Notion, or surface action items in your task manager. If you're in sales, consulting, or client services where face-to-face meetings generate follow-up work, this combo removes the manual step of moving notes from one silo to the tools where work actually happens. Zapier Plaud integration
YouTube brings conversational AI to TV screens. The feature (previously mobile-only) lets you click an "Ask" button while watching and ask questions about the video without pausing. Recipe ingredients, song lyrics, actor bios, all answered instantly. If you create YouTube content for training, onboarding, or product demos, understand that your audience can now interrogate your videos in real time. Make sure your content is complete and accurate, because viewers will fact-check you mid-watch. Available in English, Hindi, Spanish, Portuguese, and Korean for users 18+. YouTube TV AI
Reload launches Epic, an AI agent with shared memory. If you're using multiple AI coding agents (like Cursor, Windsurf, or similar), they often lose context as your project evolves. Epic acts as an architect that continuously defines product requirements, maintains shared system understanding, and reminds agents what they're building and why. It's designed for teams where different engineers use different AI tools on the same project. $2.275M seed round led by Anthemis. If you're building software with AI assistance and finding your codebase drifting from original intent, this is designed to solve that problem. Reload AI agent
Mirai raises $10M to speed up on-device AI. London-based startup building an inference engine that runs AI models 37% faster on Apple Silicon without touching model weights (no quality loss). The SDK integrates with a few lines of code. Focus is on text and voice, with vision support coming later. If you're building consumer apps that run AI locally (privacy-focused, offline-capable, cost-optimized), this could cut latency and improve user experience. Seed round led by Uncork Capital. Android support planned. Mirai on-device AI
Reddit testing AI shopping search. Small group of U.S. users now see product carousels with pricing and where-to-buy links when searching for things like "best noise-canceling headphones." Products are pulled from actual Reddit discussions, so recommendations reflect real user opinions. For businesses, this means Reddit threads are now direct purchase pathways. If your products are discussed on Reddit, monitor those conversations closely and make sure purchase paths are clear. Reddit's weekly search users grew from 60M to 80M (30% increase) over the past year. Reddit AI shopping
AI tools are replacing tasks, not roles. At Web Summit Qatar, CEOs from Read AI and Lucidya explained how customer support teams are using AI to handle volume while agents shift to supervisory roles, relationship-building, and business development. Read AI's customer service team is five people serving millions of monthly users. The pattern: AI handles the repetitive work (note-taking, ticket triage), humans handle judgment calls and strategy. If you're worried AI will eliminate positions, focus on repositioning your team for higher-value work that AI can't do. AI replacing tasks
Accenture mandates AI use for leadership promotions. Associate directors and senior managers must show "regular adoption" of AI tools to advance to leadership roles. 550,000 employees already reskilled on generative AI fundamentals. The message: if you're in professional services or management consulting, AI fluency is no longer optional. Start using it now, or your career ceiling drops. The policy applies globally except in 12 European countries and U.S. government contract divisions. Accenture AI mandate
Open source projects drowning in AI-generated bad code. VLC and Blender report declining quality in merge requests as AI coding tools flood repos with low-quality submissions. cURL halted its bug bounty program after being overwhelmed by AI-generated security reports with no substance. Developer Mitchell Hashimoto launched a system to limit GitHub contributions to "vouched" users only. If you maintain open source projects or accept outside contributions, expect to implement quality gates. The easy-to-use nature of AI coding tools has eliminated natural barriers to entry. Open source AI problems
Gemini 3.1 Pro pricing: $2/million input tokens, $12/million output (under 200K tokens). That's less than half the price of Claude Opus 4.6. If you're running high-volume AI workloads, test this model against your current provider and run the cost comparison. Gemini pricing
OpenAI nearing $100B raise at $850B+ valuation. Pre-money value: $730B. Amazon ($50B), SoftBank ($30B), Nvidia ($20B), and Microsoft leading the round. VC firms and sovereign wealth funds closing later. The company is testing ads in ChatGPT for free users to accelerate path to profitability. If you're a free ChatGPT user, expect your experience to change soon. If you're paying for Plus or Enterprise, you're safe for now. OpenAI fundraising
India allocates $1.1B for state-backed AI VC fund. Investing in AI and advanced manufacturing startups across the country. Blackstone invested $600M in AI startup Neysa (majority stake). Adani committed $100B for AI data centers by 2035. Reliance pledged $110B over seven years for gigawatt-scale data centers. If you're building AI tools or infrastructure and willing to expand to India, this is where the money is flowing. India AI investments
Anthropic and OpenAI both open India offices. Anthropic in Bengaluru, OpenAI in Bengaluru and Mumbai. India is now the second-largest market for both Claude and ChatGPT (100M+ weekly ChatGPT users). Anthropic partnering with Infosys to deploy Claude in enterprises. OpenAI partnering with Tata Group for 100 megawatts of AI compute. For SMBs in India or serving Indian markets, localized AI support is coming. Expect better latency, regional pricing, and local language improvements. Anthropic India | OpenAI India
Prompt injection attacks are escalating. A hacker exploited the Cline AI coding tool to automatically install OpenClaw on users' computers via prompt injection. The vulnerability was privately reported weeks earlier but only fixed after public disclosure. If you're building AI agents or tools that execute code on behalf of users, prompt injection defenses are critical. OpenAI introduced Lockdown Mode for ChatGPT to prevent data leaks if hijacked. Prompt injection attack
Microsoft proposes standards for proving what's real online. New blueprint for verifying digital content using provenance tracking, watermarking, and digital signatures. Evaluated 60 combinations of authentication methods. California's AI Transparency Act takes effect in August. Microsoft hasn't committed to implementing its own recommendations across Copilot, Azure, or LinkedIn. If you publish content online or operate a platform, verification standards are coming. Start planning now. Microsoft verification standards
Test Gemini 3.1 Pro for cost savings. If you're using Claude Opus 4.6 or GPT-4 for high-volume tasks (summarization, classification, data extraction), run parallel tests with Gemini 3.1 Pro. Same capability, half the cost. Use Google AI Studio (free tier) for testing before committing to production. Google AI Studio
Automate meeting follow-ups with Plaud + Zapier. If you're in sales or consulting and spend time manually transferring meeting notes to your CRM, set up a Zap that auto-archives Plaud transcripts to Google Drive, pushes summaries to Notion, and creates tasks in your project management tool. One-time setup, recurring time savings. Plaud Zapier templates
The AI industry is shifting from raising money to making money. OpenAI testing ads, nearing $100B raise at $850B valuation. Anthropic and OpenAI reportedly considering IPOs this year. India emerged as the second-largest AI market this week, with $200B+ in infrastructure commitments and both major labs opening local offices. For small businesses, this means: (1) pricing competition is heating up (Gemini 3.1 Pro undercutting Claude/GPT), (2) AI tools are moving from "nice to have" to mandatory in professional settings (see Accenture's promotion policy), and (3) security risks are escalating (prompt injection attacks, AI-generated code flooding open source). Watch for more aggressive pricing moves as models compete, and start building AI fluency into your team's core skills. The companies that treat AI as optional are the ones falling behind.