Get tomorrow's brief in your inbox
Today: GitHub Copilot switched to token-based billing, spiking some developers' costs from $29 to $750 per month. Zendesk launched $100M in free AI tools for startups. Browser makers are racing to build AI agents that can browse, shop, and write code on your behalf.
GitHub Copilot just made a dramatic shift that has developers either defending it or canceling their subscriptions. Starting June 1, Microsoft is moving from flat-rate subscriptions to token-based billing, where you pay based on how many AI-generated tokens you consume while coding. Some developers reported their bills jumping from $29/month to $750, while others claimed increases from $50 to $3,000. Read more
The controversy splits into two camps. Critics say Microsoft encouraged heavy usage and is now pulling the rug out. Defenders argue that anyone whose bill spiked this much is "vibe-coding" with no real development knowledge, blindly accepting bloated AI suggestions instead of using it as a focused tool. The truth likely sits in the middle: Microsoft was subsidizing usage at unsustainable rates, and now the real economics are showing up.
For small business owners using AI coding assistants, this is a wake-up call. Free tiers and low flat rates often hide the true cost of AI tools. Before you commit workflows to an AI service, check whether pricing is based on usage rather than a fixed subscription. If you're building a product that relies on AI APIs, those costs can balloon as you scale. The lesson here is to test with real workloads and track token consumption before you're locked in.
Zendesk for Startups: Zendesk announced a $100 million commitment to provide startups with up to two years of free access to its full AI-powered customer support platform. The package includes AI agents for automating support tickets, dedicated onboarding, and partner benefits from AWS and GitHub. Over 50,000 startups already use Zendesk's tools, including companies like ElevenLabs and Canva. Learn more
If you're running a lean operation where every team member's time matters, automating customer support from day one can free you to focus on building your product instead of answering the same questions. The catch is the learning curve. Not every founder or team member will be comfortable implementing AI-driven workflows, and setup requires time that's often scarce at early stages. Still, two years free is a serious runway extension for bootstrapped companies.
AI-Powered Browsers: The browser wars are heating up with AI at the center. Perplexity launched Comet, an AI browser that acts as a chatbot-based search engine, summarizing emails and sending calendar invites. It's currently limited to users on Perplexity's $200/month Max plan. The Browser Company introduced Dia, an invite-only AI browser that can recall every website you've visited and help with research and tasks. Opera is building Neon, which can research, shop, and write code even while you're offline (pricing not yet announced). OpenAI's Atlas browser lets you ask ChatGPT about search results and browse websites inside the chatbot itself. Full roundup
These tools promise to handle repetitive web tasks, but they're all subscription-based and mostly in closed beta. If you spend hours each week doing manual research or data entry across web apps, joining a waitlist now could pay off in six months. For most small businesses, wait until these tools prove their value before committing.
Google Gemini Spark: Google launched Gemini Spark, a 24/7 AI assistant that runs in the cloud so you can close your laptop while it works. It integrates with Gmail, Calendar, Docs, Sheets, and Slides to handle tasks like scanning your inbox, drafting expense reports, or planning your weekend. Early testing showed mixed results: it found real savings on drugstore deals but suggested invalid promo codes, and it can't use Google Keep for simple packing lists. Read the hands-on review
Gemini Spark is better suited for work tasks than personal productivity, and Google struggles to pitch clear use cases that feel essential rather than optional. If your business already lives in Google Workspace, it's worth testing for tasks like email triage or calendar summarization. Just don't expect it to replace your brain yet.
IBM Red Hat AI Inference: IBM unveiled Red Hat AI Inference and Red Hat OpenShift Virtualization Service, both aimed at small businesses that want to add AI to their applications without hiring AI specialists. The AI Inference service lets you deploy AI models directly into your apps for tasks like automating customer inquiries or analyzing sales trends. The Virtualization Service lets you run traditional and cloud-native apps on a single platform, reducing infrastructure complexity. Details here
This is enterprise-grade tech packaged for smaller teams, but it still requires a culture shift and staff training. If you're already using Red Hat OpenShift, these additions can streamline operations. If you're not, the learning curve and time investment may outweigh the benefits unless you have clear automation goals and the bandwidth to implement them.
Betting on the Old-School Web: Craig Campbell walked away from a "blank check" VC offer to build AI tools and instead created Past Maps, a website that overlays historical maps with modern ones. He built it to support his metal detection hobby and shared it on Reddit, where demand grew organically. Traffic climbed from 20,000 to 300,000 monthly active users in three years, fueled entirely by Google Search. He charges $9/week or $52/year for full access, avoiding the ad-tech treadmill. Read the story
Campbell's revenue matches what he earned as a mid-level engineer at Meta, not the VC-backed windfall he could have chased. But he owns the business, serves a real niche, and doesn't answer to investors. For small business owners, this is a reminder that sustainable revenue from a focused audience beats chasing hype. If you have deep knowledge of a niche and can solve a real problem, subscription models protect you from ad revenue swings and platform algorithm changes.
AI Scams on TikTok: Dropshippers are using AI-generated influencers, particularly fake Black women, to sell mass-produced Shein products at marked-up prices on TikTok. These AI avatars cry on camera, claim to hand-make items like belt buckles, and use guilt-based appeals to drive sales. Researchers are finding up to 100 new AI-generated retail scam accounts every day. One account, "Aliyahsbuckles," has 40,000 followers and sells $9 Shein belt buckles for $40. Full investigation
This is a warning for any business selling online: AI-generated content is cheap, scalable, and deceptive. Platforms don't police this well, and customers can't always spot the fakes. If you're competing with dropshippers, transparency is your advantage. Show your face, show your process, and build trust that AI grifters can't fake. If you're buying ads or working with influencers, verify they're real before handing over money.
Anthropic's Revenue Model: Anthropic calculates "run-rate revenue" by taking the last 28 days of consumption-based sales, multiplying by 13, then adding 12 months of subscription revenue. Source
This is useful context for understanding how AI companies report growth. If you're building on top of Anthropic's APIs or any consumption-based AI service, remember that your usage directly feeds their revenue projections. Budget accordingly.
Perplexity Comet Pricing: Comet is only available to users on Perplexity's $200/month Max plan, with a waitlist for everyone else. This is prohibitively expensive for most small businesses, especially when the feature set is still unproven.
Meta's AI Pendant: Meta is developing an AI-powered pendant that users can wear to record conversations, building on its acquisition of Limitless at the end of 2025. Testing is expected to start within the next year. The company is also planning to expand its AI glasses lineup and launch a "Wearables for Work" business subscription. Report
AI wearables have flopped so far due to privacy concerns and limited utility. Meta is betting that workplace use cases will fare better, but this feels like throwing hardware at a problem that software hasn't solved yet. If you're considering AI wearables for your team, wait for real-world reviews and clear ROI data before buying in.
SoftBank's French Data Centers: SoftBank announced plans to invest up to €75 billion (roughly $87 billion) to build data centers in France, aiming for 5 gigawatts of capacity by 2031. This is SoftBank's largest AI infrastructure investment in Europe. Details
Massive data center investments signal where AI infrastructure is heading, but they also raise questions about environmental impact and utility costs. For businesses relying on cloud AI services, expect capacity to grow but also expect regulatory scrutiny and potential cost volatility as energy demands spike.
Anthropic vs. OpenAI: An Inc.com piece argues that Anthropic's rise over OpenAI is a case study in emotional intelligence, though it offers no concrete details beyond the headline. Read the tease
This is content-farm filler, not actionable insight. Skip it.
Audit Your AI Tool Costs: If you're using GitHub Copilot, Claude, ChatGPT, or any token-based AI service, track your actual usage before June bills arrive. Many tools offer usage dashboards. Export your data, calculate your monthly burn rate, and compare it to alternatives. You may find that a different tool or a manual workflow is cheaper for low-frequency tasks.
Test Zendesk for Startups: If you're pre-revenue or early-stage and drowning in support tickets, apply for Zendesk's startup program. Two years of free AI-powered support automation could be the difference between you answering emails all day and you building features that drive revenue.
The GitHub Copilot pricing shift reveals a broader truth: AI subsidies are ending. Tools that seemed too cheap to be true were exactly that. As companies move from land-grab mode to profitability mode, expect more usage-based pricing, tiered plans, and surprise bills. The browser wars are heating up with AI agents that promise to handle tedious web tasks, but most are still invite-only or expensive. If you're betting on AI tools to run your business, build in cost buffers and have fallback plans. The hype is real, but so is the bill.