Get tomorrow's brief in your inbox
Today: OpenAI is hiring for family-focused AI products as ChatGPT usage among parents jumps 50% year-over-year. Tech layoffs hit 140,000 in 2026 as companies cite AI automation in 23% of cuts. Meanwhile, chip and data center executives report supply constraints with demand booked out five years.
Tech Companies Replace Workers With AI While Demand for AI Infrastructure Outpaces Supply
The AI transition is creating a stark divide. Amazon has laid off 57,000 corporate workers since 2022 (16% of its corporate workforce), with CEO Andy Jassy warning that AI "will reduce our total corporate workforce" through efficiency gains. Across the tech sector, 140,000 U.S. employees lost jobs in 2026 so far, with AI cited as the reason in 23% of layoff announcements. Some Amazon employees described working on AI projects only to find themselves replaced by the technology they helped build.
At the same time, AI infrastructure companies report almost unlimited demand. Lumentum, which makes optical connectivity for data centers, is sold out for the next five years. Cerebras and Nebius both report demand far exceeding capacity. Pat Gelsinger, former Intel CEO, told CNBC that AI demand is "almost unlimited" with energy availability as the only real constraint.
What this means for small businesses: If you're considering AI tools, availability and pricing should stabilize as supply catches up. But if you're in professional services or knowledge work, automation pressure is real. The companies thriving are those using AI to augment their teams, not replace them. Start identifying which parts of your work AI can handle so you control the transition instead of reacting to it.
Amazon layoffs | AI infrastructure demand
OpenAI Expands Into Family and Elder Care Products
OpenAI is hiring a dedicated product manager in San Francisco to build experiences for families, caregivers, and older adults. The move comes as ChatGPT's user base shifts older: users 35+ grew from 26% to 31% globally in the past year, while the 18-24 age group dropped from 34% to 29%. Nearly one in four U.S. smartphone users who are parents now use ChatGPT, up from 16% a year earlier.
The company has added parental controls for teen accounts and a "Trusted Contact" feature that can alert family members in potential self-harm situations. New research from the Family Online Safety Institute found parents underestimate their children's AI use: 38% of kids reported using generative AI in the past week versus only 27% of parents who thought their child had used it.
If you run a service business targeting families (tutoring, coaching, elder care), watch this space. OpenAI is building guardrails and age-appropriate experiences that could make AI assistants safer for family use. Consider how AI tools might fit into your service delivery, but verify they have proper content controls before recommending them to clients with children.
sqlite-utils 4.1 Adds Code-Based Row Generation
Simon Willison released sqlite-utils 4.1 with a new --code option that lets you generate database rows using Python code instead of importing from files. The tool now accepts --type overrides for columns (useful for ZIP codes that should be text, not integers), can infer primary keys for upserts, and supports toggling SQLite's strict mode on existing tables.
This is useful if you're building lightweight data pipelines or need to manipulate SQLite databases (common in local-first apps and data analysis). The new code-based insertion is particularly handy for generating test data or transforming inputs on the fly. Free and open source.
Enterprise AI Spending Shifts From "Tokenmaxxing" to "Valuemaxxing"
Enterprises are changing how they buy AI. The industry is moving from "tokenmaxxing" (use as much AI as possible regardless of cost or result) to what executives are calling "valuemaxxing" (focus on ROI and cost-effective models). Companies are questioning whether expensive frontier models from OpenAI and Anthropic are necessary when open-source alternatives from DeepSeek and Alibaba cost significantly less.
This doesn't mean AI demand is slowing. Data center operators report extraordinary demand and multi-year backlogs. What's changing is enterprises are getting smarter about which model to use for which task. Use an expensive reasoning model for complex analysis, but switch to a cheaper model for summarization or classification.
For small businesses, this is good news. You don't need the most expensive AI subscription to get value. Test free tiers and open-source models first. Save premium models for tasks where quality really matters (client-facing work, complex decisions). Many businesses are wasting money on enterprise AI subscriptions when a $20/month ChatGPT Plus account would cover their actual needs.
Meta Pauses AI Image Feature After Privacy Backlash
Meta introduced Muse Image, an AI feature that let Instagram users create AI-generated images from other people's posted photos without permission. Three days later, Meta paused the feature after privacy complaints. The company admitted "we missed the mark."
This follows a pattern of AI companies launching features that use user content for training or generation without clear consent. For businesses using social media for marketing, the lesson is clear: assume anything you post publicly could be scraped for AI training. If you have proprietary product images, branding, or visual IP, watermark them and monitor for AI-generated copies.
China Bets on Non-Invasive Brain-Computer Interfaces
While Elon Musk's Neuralink drills into skulls, China's BrainCo is building non-invasive brain-computer interfaces using wearable devices. The company makes FDA-approved bionic hands that read neural signals and wearable sleep aids using low-intensity electrical pulses. BrainCo raised $280 million in funding and plans to expand from medical devices into consumer electronics, then license its platform to other companies.
China designated BCI as a strategic "future industry" in its Five-Year Plan. The technology isn't ready for mainstream business use yet, but it's worth watching if you're in health tech, accessibility, or consumer electronics. The non-invasive approach has lower barriers to adoption than implants.
Fix AI Implementation Failures by Fixing Management First
If your AI rollout isn't working, the problem might not be the tools. Multiple sources this week highlighted that AI tools fail when organizations have broken processes or poor leadership. Better AI can't fix bad management, unclear workflows, or teams that don't trust leadership.
Before buying another AI subscription, map out your actual workflow. Where are the bottlenecks? What tasks take the most time? What information gets lost in handoffs? Use AI to solve specific, well-defined problems in functional processes. Don't use it to paper over organizational dysfunction.
AI rollout and corporate culture | AI in healthcare adding time instead of saving it
The AI market is maturing. Infrastructure demand remains strong with multi-year backlogs, but enterprise buyers are getting more sophisticated about cost and ROI. Companies are still investing heavily in AI, but they're questioning which models they actually need. For small businesses, this means more options at better prices as the market sorts out which use cases need premium models versus commodity AI. Watch for continued job displacement in corporate roles alongside growing demand for people who can effectively use AI tools.