In this issue
Perplexity just rolled out a service that turns a dedicated Mac Mini into a local, always-on digital employee. Instead of running one-off prompts, users can now orchestrate multiple models directly against their own files to execute background tasks.
Topics of the day:
Perplexity transforms the Mac Mini into a local digital worker
Anthropic launches Opus 4.7 for reliable, autonomous task execution
PwC reveals only 20% of companies capture 74% of AI's value
Snap cuts 1,000 jobs as AI drives new coding efficiency
Meta launches its first major AI model after the $14B Wang hire
Curated reads on agents, Claude scheduling, the Stanford AI Index, and more
The Shortlist: Anthropic passes OpenAI in revenue, exec exodus at OpenAI, Claude Design, and more
Perplexity reimagines the personal computer
What's happening: Perplexity just launched a service that transforms a dedicated Mac Mini into an always-on digital worker. Available to Max users, it orchestrates 20 different models across your local files and applications so you can control tasks from anywhere.
In practice:
You can offload heavy research tasks to this background machine while keeping your primary laptop completely free for focused work.
The system runs directly against your local files and apps, meaning it acts with full business context without requiring you to manually upload sensitive documents.
Operators can set up multi-step workflows that run around the clock, turning a simple piece of hardware into a tireless operational assistant.
Bottom line: This bridges the gap between basic cloud chatbots and true local automation. If you have an extra Mac Mini available, you now have a dedicated digital employee ready to scale your output.
Anthropic drops Claude Opus 4.7
What's happening: Anthropic just launched Opus 4.7, bringing notable upgrades to software engineering, high-resolution vision, and reasoning effort controls. This new model takes instructions far more literally and handles long-running, autonomous tasks with fewer interruptions.
In practice:
You can now use an "xhigh" effort level to give the model more time to catch its own logic faults on complex business problems.
The new /ultrareview slash command acts as a dedicated review session to flag code bugs and design issues before you push updates.
Because the model follows instructions strictly, use Anthropic's migration guide to retune your existing prompts and avoid unexpected outputs.
Bottom line: Opus 4.7 is a direct upgrade that makes handing off complex tasks to AI much more reliable. As you update your workflows, lean into its improved instruction-following rather than relying on loose interpretations.
Only 20% of companies capture 74% of AI's economic value
What's happening: PwC just dropped a major study surveying 1,217 senior executives across 25 sectors. The headline: the top 20% of AI-adopting companies deliver 7.2x higher AI-driven financial performance than the rest.
In practice:
The leaders focus on growth and business reinvention, not just cost-cutting. If your AI strategy is only about saving money, you are already in the wrong 80%.
Top performers are 1.9x more likely to deploy autonomous, self-optimizing AI systems rather than basic chatbots or copilots.
The gap is widening, not closing. Companies that wait another year to move beyond pilot projects are compounding the distance between themselves and the leaders.
Bottom line: This is the most quantified wake-up call of the year for operators. If you are a founder, the first question to answer is which bucket you are in, and the second is what you are doing this quarter to move.
Snap cuts 1,000 jobs as AI writes 65% of their new code
What's happening: Snap just announced 1,000 layoffs, a 16% workforce reduction driven by AI efficiency gains. The company is leaning hard into automation, with AI now writing 65% of their new code and allowing small squads to replace traditional teams. This follows a broader trend: nearly 80,000 tech workers were laid off in Q1 2026, with close to half of those positions attributed to AI.
In practice:
You can restructure large departments into small squads that use AI to automate repetitive tasks and lower your operational overhead.
Letting AI handle your boilerplate code means you ship updates faster, but always keep a human reviewing the final output.
Automating your routine customer support queries allows you to reinvest saved capital directly into personalized experiences that drive real growth.
Bottom line: The shift from relying on massive headcount to leveraging AI efficiency is not theoretical anymore. Start building automated, lean systems in your own business before your competitors figure it out.
What I read/use this week
Hermes Agent is OpenClaw's biggest threat - A deep dive into the agent framework that could challenge OpenClaw's dominance in autonomous AI.
Claude Schedules vs Routines: what's the difference - A practical breakdown of the two new automation features in Claude and when to use each.
Stanford AI Index 2026 - The definitive 423-page report on AI's state. Key stat: the US-China gap has shrunk to 2.7%.
Gemini gets a native Mac app - Google's free desktop AI app with screen sharing and Option+Space quick access.
HockeyStack raises $50M for AI revenue agent - Their AI reverse-engineers your past won deals to build a custom prospecting blueprint.
Cloudflare Email Service enters public beta - Email sending built into the Cloudflare stack, with Workers integration for developers.
Meta debuts its first major AI model after the $14 billion Wang hire
What's happening: Meta just launched its first major AI model since acquiring the talent package led by Liqian Wang. The model represents Meta's clearest signal yet that it intends to compete directly with OpenAI, Anthropic, and Google at the frontier.
In practice:
Meta's approach combines its massive user data advantage with frontier model capabilities, meaning the next generation of Instagram, WhatsApp, and Threads features will be AI-native from the ground up.
Meta has historically open-sourced its models through the Llama series. If this new model follows suit, it could become the strongest open-weight foundation model available.
The $14B talent bet shows how expensive the AI arms race has become. For startups, this raises the bar on what "state of the art" means and accelerates the timeline on commoditized AI infrastructure.
Bottom line: When the company with 3.3 billion monthly users ships a frontier model, every AI-adjacent business feels the ripple. Watch whether Meta open-sources this one, because that decision alone could reshape the competitive landscape.
The Shortlist
Anthropic surpassed OpenAI in annualized revenue at $30B ARR, with business clients spending $1M+ per year doubling from 500 to 1,000 in under two months.
OpenAI lost three executives in a single day, including its VP for Science and Sora lead, while preparing for an $852B IPO.
Anthropic launched Claude Design, a tool that turns conversational prompts into polished prototypes, slide decks, and one-pagers. Figma stock dropped 7.28% on the announcement.
Cerebras filed for a Nasdaq IPO at a $35B valuation after securing a $20B+ chip deal with OpenAI.
Allbirds pivoted away from sustainable sneakers to become a GPU-as-a-Service business, sending its stock up 582% before falling 35% the next day.
OpenAI expanded its Codex agent beyond engineering into broader knowledge work across ChatGPT, IDEs, and the desktop.
Cal.com close-sourced its product, reigniting the debate on whether AI is killing open source as a sustainable business model.
This newsletter is where I (Kwadwo) share products, articles, and links that I find useful and interesting, mostly around AI. I focus on tools and solutions that bring real value to people in everyday jobs, not just tech insiders.
