Get tomorrow's brief in your inbox
Today: AI costs dropped 8x while tech companies cut 140,000 jobs citing AI transformation. New Jersey banned AI surveillance pricing, and librarians are hosting packed workshops teaching people how to turn AI features off.
The AI cost-benefit paradox is here, and small businesses need to pay attention. AI inference costs have dropped roughly 8 times in the past year, making tools that were prohibitively expensive now accessible to teams of any size. At the same time, U.S. tech companies have slashed nearly 140,000 jobs since January, with Monday.com this week joining a list of 20+ companies explicitly citing AI as a factor in layoffs. Monday.com cut 20% of its workforce (600+ employees) as part of what it calls an "AI-driven growth strategy," though co-founder Eran Zinman insists the move "was not made to replace people with AI."
The disconnect tells you what matters: cheaper AI is not automatically better for your business unless you reinvest the savings strategically. Companies citing AI in layoffs have underperformed the Nasdaq by almost 10% in the 30 days following their announcements, according to Financial Times analysis. Meanwhile, AI-focused companies like Anthropic and OpenAI are hiring rapidly. The takeaway for small business owners: falling AI costs create an opportunity to expand capabilities, not just cut expenses. If you treat AI purely as a headcount replacement, the market (and your customers) will notice.
What to do: Audit where you're currently spending on AI tools. If costs have dropped, reinvest the difference in training, experimentation, or expanded use cases rather than banking the savings.
Ruff v0.16.0 (Python linting tool) - Astral shipped a major update to Ruff, their Python code quality tool, now enabling 413 rules by default (up from 59). The tool catches syntax errors, runtime issues, and code quality problems automatically. For developers building AI integrations or data pipelines in Python, this is a time-saver. It's free and open-source. Try it on any Python project with uvx ruff@latest check . to see what it finds. The new version is aggressive about fixes - one project saw 1,618 errors detected, 1,538 auto-fixed. Ruff v0.16.0 release
Young workers aren't AI-native by default. A new Harvard study reveals that younger employees don't automatically know how to turn AI into business results. The assumption that Gen Z and millennials will "just figure it out" is costing companies productivity. Instead of assuming ease, businesses need structured onboarding: show employees which AI tools solve which problems, provide prompt templates, and create feedback loops to measure what's working. The study suggests hands-on training beats discovery-based learning for AI adoption. Harvard study coverage
ChatGPT is sending customers to parts of your website nobody manages. AI search tools like ChatGPT are directing traffic to pages you may have forgotten exist: old product listings, outdated FAQs, archived blog posts. If those pages are broken, stale, or inconsistent with your current messaging, you're losing customers before they ever reach your actual offerings. Run a site audit, identify pages getting AI referral traffic, and either update them or redirect them to current equivalents. Read more
AI inference costs dropped 8x - spend more, not less. The cost to run AI models has fallen dramatically in the past year. Most businesses see this as a chance to cut budgets. The teams pulling ahead see it as a chance to expand: run more experiments, test more use cases, automate more workflows. If a task that cost $100/month now costs $12.50, you can either pocket $87.50 or run 8x the volume. The latter compounds into a competitive advantage. Inc.com analysis
New Jersey bans AI surveillance pricing, joining two other states. New Jersey became the third state to prohibit AI-powered "surveillance pricing" - using customer data like income, browsing history, and past purchases to set individualized grocery prices in real time. At least 20 other states are weighing similar legislation. If your business uses dynamic pricing algorithms, review whether your data inputs cross into restricted territory. Personalization is legal; discrimination is not. The line is getting clearer. NJ ban details
Librarians are hosting viral "Avoiding AI" workshops. Public libraries are seeing unprecedented turnout for workshops teaching people how to disable Apple Intelligence, Gemini, and other AI features. One librarian in Philadelphia saw 2,000+ Instagram likes for a single event post (versus the usual few dozen). Attendees share tips like appending &udm=14 to Google searches to hide AI results. The backlash is real, and it's not just from tech skeptics - it's from people frustrated by AI being "forced onto them" without clear value. If you're building AI features, make them opt-in and explain the benefit upfront. Library workshops story
AI distillation becomes a national security debate. The White House accused Chinese AI lab Moonshot of using "distillation" - training a model using outputs from advanced U.S. models like Anthropic's Fable - to rapidly catch up without the billions in R&D investment. More than 20 tech companies (Nvidia, Microsoft, Meta, Palantir) responded with a letter urging policymakers to avoid "premature restrictions" on open-weight AI models and distillation techniques, calling it "a widely used technique for model improvement." For small businesses, this matters because open-weight models (free, downloadable, customizable) may face future regulatory limits. If your stack depends on open models, monitor this closely. Distillation debate
AI data centers caused a 10-minute grid recovery delay. A fallen power line near Washington, DC, triggered 3+ gigawatts of data center load to drop off PJM's electrical grid nearly simultaneously, causing voltage spikes and flickering lights from Northern Virginia to Chicago. The incident exposed a growing infrastructure problem: data centers switching to backup power in unison can destabilize grids. For businesses relying on cloud AI services, this underscores latency and availability risks as AI infrastructure scales. Data center grid issue
Google Search without AI results: Append &udm=14 to any Google search URL to hide AI-generated summaries and get traditional link results. Useful when you need primary sources, not summaries. (Tip shared at a Philadelphia library workshop.)
Audit your AI tool spending this month. If you're paying for ChatGPT Plus, Claude Pro, or API credits, check your usage. Costs have dropped significantly - you may be overpaying for a legacy plan or underutilizing what you have.
This week's stories reveal a maturing AI market: costs are falling, backlash is growing, and the gap between AI hype and AI results is widening. Companies that treat AI as a cost-cutting tool are underperforming. Companies that treat it as a capability expander are hiring. For small businesses, the message is clear: cheaper AI is only valuable if you use it to do more, not less. Watch for AI pricing/privacy regulations expanding beyond New Jersey, and expect more scrutiny on how AI tools use customer data.