In partnership with:
In Today’s Issue:
🎤 Gemini Audio gets a major upgrade
📉 The Labor Revolution is Here: Copywriters report an income collapse
🥇 Zoom shockingly beats the competition, hitting a new State-of-the-Art score of 48.1% on the brutal Humanity’s Last Exam benchmark.
🕰️ TIME names the "Architects of AI" its 2025 Person of the Year
✨ And more AI goodness…
Dear Readers,
What if the biggest AI story isn't the technology itself, but the disconnect between the people building it and the people losing their jobs to it? Today, we look at TIME’s decision to crown the "Architects of AI" as Person of the Year, celebrating the trillion-dollar boom. But immediately after, we dive into the reality on the ground: a brutal "market reset" for copywriters and creatives who are finding that "AI assistance" often turns into replacement.
We also have major technical updates: Google is making voice agents much better at following instructions, and Zoom (yes, that Zoom) just beat the world’s best models on a reasoning benchmark by betting on federated small models rather than one giant monolith.
All the best,
🎼 Gemini Audio Gets Major Upgrade
Google just upgraded Gemini 2.5 Flash Native Audio for live voice agents. It is now significantly better at multi-step “do-this-then-that” workflows, hitting 71.5% on ComplexFuncBench Audio, and is more obedient to developer instructions (90% adherence, up from 84%). The update is rolling out across AI Studio, Vertex AI, Gemini Live, and Search Live.
📚 TIME Crowns The AI Architects
TIME named the “Architects of AI” its 2025 Person of the Year, arguing that 2025 was the moment AI shifted from “debate” to an all-out deployment sprint. The cover is fronted by figures like Nvidia CEO Jensen Huang, whose company rode the chip boom to become the world’s first $5T company. Nevertheless, the selection seems strange, placing CEOs like Zuckerberg and Huang alongside scientists like Hassabis gives the impression of arbitrariness rather than a cohesive group.
📺 Zoom Tops Humanity’s Last Exam
This came as a surprise: Zoom (yes, Zoom!) says it just hit a new SOTA 48.1% on the Humanity’s Last Exam (HLE) full-set, beating the prior best of 45.8% held by Gemini 3 Pro with tools. That’s a +2.3 point jump on one of the toughest expert-level reasoning benchmarks.
The key idea is federated AI: orchestrating Zoom’s small models with a mix of open and closed models (plus a “Z-scorer” and an explore–verify–federate workflow) instead of betting on one monolith. This could translate into reliable agentic features in AI Companion 3.0, like better summaries, retrieval, and multi-step workflow automation.
The arrival of AGI | Shane Legg (co-founder of DeepMind)
Critics of TIME: the revolution in the labor market is here!
The Takeaway
👉 Copywriters describe a repeated pattern: AI arrives as “assistance,” then the job shifts to editing AI output, and many end up laid off or losing clients once the output is deemed “good enough.
👉 Several accounts point to a market reset: clients accept lower quality because it’s cheaper, pushing rates down and turning craft work into template-driven commodity content.
👉 The human cost is front-and-center: people report closed businesses, income collapse, and career pivots, with a sense of being devalued and dehumanized.
👉 This echoes broader reporting about white-collar disruption: AI isn’t just changing tools, it’s reshaping who gets paid, how much, and for what.
AI didn’t just “assist” copywriters, it retrained the market to accept cheaper, blander language. In Brian Merchant’s latest AI Killed My Job installment, copywriters describe a familiar arc: first you’re told AI will “speed you up,” then your role quietly becomes editing prompts, and finally you’re replaced because the bot is “good enough.”
One freelancer says their entire client queue vanished after a sale, replaced by Mad-Libs-style templates and AI. Another describes a thriving agency collapsing from hundreds of thousands in annual revenue to almost nothing. A medical writer reports hours plunging from near full-time to a handful per month, while “review” work gets pushed to higher-credentialed staff for less pay. And multiple writers describe the same gut-punch: clients building custom GPTs that appear to be trained on the very writing they paid humans for.
The AI community should read this as a product lesson: when you automate the “first draft,” you also automate the value perception, and the labor market resets fast. We have already arrived deep in the labor revolution.
(image: The impact of artificial intelligence on the labor market; czech national bank)
Why it matters: This is what “AI adoption” looks like on the ground, not instant replacement, but a slow squeeze from creator → editor → expendable cost. If we don’t design for provenance, consent, and quality incentives, we’ll scale content while shrinking careers.
Sources:
🔗 https://www.bloodinthemachine.com/p/i-was-forced-to-use-ai-until-the?r=n1p8t
🔗 https://www.theguardian.com/technology/2025/may/31/the-workers-who-lost-their-jobs-to-ai-chatgpt
🔗 https://www.cnb.cz/en/about_cnb/cnblog/The-impact-of-artificial-intelligence-on-the-labour-market/
Mino: Build on 95% of the Web Without APIs
Most valuable data doesn’t live behind APIs. It’s hidden behind logins, forms, booking flows, and complex navigation that traditional scrapers can’t handle. Browser-based AI agents are slow and expensive, guessing where to click on every run.
Mino is a web agents API that navigates websites intelligently and returns structured JSON. It learns how sites work once, then executes fast, deterministic workflows, getting cheaper and more reliable over time. Run tasks across multiple sites in parallel, handle JavaScript, modals, and forms, and extract data that only appears mid-workflow.
Start with one API call.
SpaceX’s $800B Accounting Ripple
SpaceX just reset the private-market ceiling. An insider tender offer priced at $421 per share implies roughly an $800B valuation. This isn’t new money flowing into the company, it’s existing employees and early investors selling shares, with buyers reportedly able to take up to about $2.56B in stock. But one “secondary” deal can still move mountains: it effectively re-marks what a SpaceX share is worth today.
That matters for Alphabet. Under U.S. accounting rules, when a private holding has an observable price change, companies can record an unrealized gain even without selling. Alphabet’s Q1 2025 results already included an $8B paper gain tied to a revaluation of a private-company stake widely understood to be SpaceX. A jump from ~$350B to ~$800B could tee up another boost in “other income.”
For the AI community, this is a reminder that infrastructure is the moat: launch and Starlink connectivity are getting AI-sized multiples. If capital keeps rewarding the pipes, not just the models, what does that unlock next? Space-based compute, global edge AI, or something we haven’t named yet?
The best marketing ideas come from marketers who live it. That’s what The Marketing Millennials delivers: real insights, fresh takes, and no fluff. Written by Daniel Murray, a marketer who knows what works, this newsletter cuts through the noise so you can stop guessing and start winning. Subscribe and level up your marketing game.
