Another Wednesday, another selection of AI news and resources to help you become more AI native. This week:
Anthropic ships Opus 4.8 with 1,000 parallel coding agents and a $965B valuation
NVIDIA and Microsoft turn the Windows PC into an agent machine
GitHub Copilot becomes a desktop app that runs every major model
The 20-minute Claude setup that sounds like you
Next Gen
Your individual contributors just got promoted to manager.
Whats happening: Tech layoffs hit 142,000 in the first five months of 2026, up 33% year over year, even as the same companies pour roughly $700 billion into AI infrastructure. Salesforce's engineering headcount has been flat at 15,000 for two years. Cognition, the maker of the Devin coding agent, just raised $1 billion at a $26 billion valuation on $492 million of revenue.
Why it matters: The doom read is that AI is replacing people. The accurate read is that AI is amplifying the people who stayed. Headcount stops growing because each remaining person now operates a team of agents. Marc Benioff said it directly on the May 28 Salesforce earnings call: "We're not hiring more engineers" because "we've been using AI to create more efficiencies for our engineers." The org chart didn't shrink. The leverage curve got steeper.
In the wild: At Anthropic's Dynamic Workflows launch this week, the showcase was the Bun runtime ported from Zig to Rust in 11 days. That was 750,000 new lines of code with a 99.8% pass rate on the existing test suite. The same project used to take a full engineering team an entire quarter. Now it's one engineer pointing the model and reviewing the output.
Looking ahead: The career split is no longer technical versus non-technical. It is the people who direct agents versus the people who wait for instructions. Pick one role on your team this quarter, including yourself, and treat that human as a manager of three to five agents starting Monday. Define the goal, hand off the grunt work, and review on a schedule. The companies that pull this off in 2026 will look the same on the org chart but ship like ones four times their size.
Need help finding where your company is losing money?
I can use AI to save your company money and build a plan to get it back. If I can’t find any meaningful savings, you don’t pay.
AI First
The 20-Minute Claude Setup That Sounds Like You
Stop using Claude like a search box and start using it as a writing partner that already has your voice.
Most people type a question and copy the answer. The real leverage is a one-time 20-minute setup that turns Claude into a collaborator who knows how you sound.
Open Claude and create a Project for the writing you do most (emails, decks, memos).
In the Project's custom instructions, paste three to five samples of your past writing and describe your voice in three sentences.
Add the audience profile, your usual goals, and any words or phrases you never use.
Inside the Project, start every new draft as a chat so Claude carries the voice forward.
Save the prompts that produce the best output as Project files for next time.
Refine the instructions weekly based on what you had to edit by hand.
Pro tip: Paste in your three worst recent drafts and ask Claude to identify the patterns you struggle with. Bake the fixes into the Project instructions so you stop repeating them.
AI News
Anthropic ships Opus 4.8 with 1,000 parallel coding agents, and a $965B valuation to match.
41 days after Opus 4.7, Anthropic released Opus 4.8 at the same $5/$25 per million tokens, plus Dynamic Workflows in Claude Code that orchestrates up to 1,000 parallel subagents per session. The proof point: the Bun runtime was ported from Zig to Rust in 11 days, producing 750,000 lines of code with a 99.8% pass rate on the existing test suite. The same week, Anthropic closed a $65 billion Series H at a $965 billion post-money valuation and confidentially filed S-1 paperwork.
NVIDIA and Microsoft turn the Windows PC into an agent machine.
At Microsoft Build 2026, NVIDIA and Microsoft unveiled RTX Spark, a new Windows platform built to run AI agents locally instead of in the cloud. The specs: 1 petaflop of AI compute, 128GB of unified memory, and the ability to run 120-billion-parameter models on-device. NVIDIA's OpenShell runtime and Microsoft's new Windows agent security primitives provide the safety layer. RTX Spark laptops and desktops from Surface, ASUS, Dell, HP, Lenovo, and MSI are expected this fall.
GitHub Copilot becomes a desktop app that runs every major model.
Microsoft used Build 2026 to ship a standalone GitHub Copilot desktop app, its direct answer to Codex and Claude Code, with native access to OpenAI, Anthropic, and Google models from one harness. Microsoft also released Aion 1.0 Instruct and Aion 1.0 Plan, two local Windows models for on-device agent workflows, plus Agent 365 and Foundry IQ for enterprise data and security. The keynote was less a product launch and more a blueprint for the agent-first operating system.
Reply to this email if you have any questions or further content you want covered.
Talk soon,
Cam