In partnership with
Sign Up | Advertise | Dispatches | Consulting
Hey Explorers!
The model layer is no longer the bottleneck. This week, every signal pointed somewhere else: the human layer.
Courts, awards bodies, the White House, and ordinary companies all got dragged into answering questions they hoped to skip.
Here's the signal that matters 👇🏼
🗞️ TECH & TRENDS | The Readiness Gap
📢 Meta Just Opened Your Ad Account to Claude and ChatGPT Meta launched Ads AI Connectors in open beta, letting Claude, ChatGPT, and any MCP-compatible agent create campaigns, manage catalogs, and edit ad sets directly through natural language. No developer credentials. No API keys. Write access from day one, which is more than Amazon or Google's MCP servers offered at launch. If you run paid social, your workflow just changed. The question is who in your org becomes the prompt engineer for spend decisions. (🔗 Link)
🪞 Daniel Miessler Said the Quiet Part Loud: Most Companies Can't Use AI Even If They Want To After consulting for the world's biggest companies and hundreds of startups, Miessler argues the AI readiness problem isn't technical. It's that most companies cannot describe what they actually do. Goals shift quarterly. Workflows are tribal knowledge. Strategy slides get rewritten before they're executed. AI is useless inside that chaos. The companies AI is helping are the same ones that already had their act together. Everyone else is buying tools that just help them flail more impressively. (🔗 Link)
🏛️ The White House Quietly Wants AI Safety Tests Back Per Axios, the Office of the National Cyber Director hosted two meetings last week with tech companies and trade groups to discuss security frameworks for advanced AI, with the Pentagon potentially leading safety testing for federal, state, and local deployments. This is the same administration that revoked Biden's AI executive order on Day 1. Anthropic's Mythos changed the math. When a model finds zero-days in every major OS and browser, even the deregulators start drafting executive orders. (🔗 Link)
🧠 6% of Claude Conversations Are People Asking AI for Life Advice Anthropic analyzed a million conversations and found roughly 6% were users seeking personal guidance. Should I take the job. How do I talk to my crush. Should I move. Three quarters of those conversations clustered in four domains: health, career, relationships, money. Anthropic also caught itself being sycophantic 25% of the time on relationship questions, and retrained Opus 4.7 specifically to push back. The brilliant friend at the kitchen table is now a category. Your customers are already in it. (🔗 Link)
🎬 The Oscars Drew a Line. AI Actors and AI Scripts Don't Qualify. The Academy ruled that performances must be "demonstrably performed by humans with their consent" and screenplays must be "human-authored" to be eligible for awards. Hollywood is the latest institution to gate its top status symbol behind verified human work. The bet is that the prestige of "made by humans" will hold its value. The counter-bet is that audiences in five years will not care, and the Oscars will quietly walk this back. Pick your side. (🔗 Link)
⚖️ A Chinese Court Just Ruled You Can't Fire Workers to Replace Them with AI A Hangzhou court ordered a tech firm to pay roughly $38,000 in compensation after it tried to demote a quality assurance supervisor with a 40% pay cut, then fired him when he refused, citing AI replacement. The court ruled that adopting AI is a strategic business choice, not an unforeseeable change in objective circumstances, so it does not qualify as legal grounds for termination. Beijing courts ruled the same way on a separate case in December. Picture this precedent landing in California or the EU. Every "we're restructuring around AI" press release becomes a litigation event. (🔗 Link)
💬 Shapes Raised $8M to Put AI Characters Inside Your Group Chats Shapes is Discord with AI characters living alongside humans in shared conversations. 400,000 monthly users, 3 million AI personas already created, thousands of users spending two to four hours a day in the app. The pitch is that group chats die because nobody wants to send the first message, and AI agents fix that by always responding and initiating. Lightspeed led the seed round. Whether or not you think this is healthy, it is where a chunk of Gen Z is already living. (🔗 Link)
📜 Someone Built an LLM Trained Only on Data From Before 1930 Talkie is a 13B-parameter "vintage LLM" trained exclusively on pre-1930 public domain material, designed to emulate the perspective of that era. Two motivations: dodge the copyright fight, and explore whether a model grounded in a specific historical moment could predict what came after. Demis Hassabis once asked whether a model trained on data through 1911 could discover general relativity. Talkie can't yet. It is currently inventing cricket matches. But the framing is the interesting part. Most AI debates assume more data is better. This argues that the right data, scoped to the right moment, might be a different kind of useful. (🔗 Link)
Every story above is the same story. The model layer stopped being the bottleneck. Your operation is. The companies pulling ahead aren't the ones with the best AI. They're the ones who can describe what they do clearly enough that AI can actually plug in.
The honest question is whether your leadership team has that clarity. Most don't. That's the gap we close.
Purple Horizons is your AI sherpa for the CEO and executive leadership team. We sit at the table, cut through the noise, and help you decide what to build, what to kill, and where AI actually compounds your edge.
Signal over noise. Always.
Ralph & Gianni
The ops hire that onboards in 30 seconds.
Viktor is an AI coworker that lives in Slack, right where your team already works.
Message Viktor like a teammate: "pull last quarter's revenue by channel," or "build a dashboard for our board meeting."
Viktor connects to your tools, does the work, and delivers the actual report, spreadsheet, or dashboard. Not a summary. The real thing.
There’s no new software to adopt and no one to train.
Most teams start with one task. Within a week, Viktor is handling half of their ops.
📅 LET’S HANG OUT | Upcoming Events
📍 5/12 Cafecito & Pastelitos @ Chug’s Diner
📍 6/2 South Miami Tech Tuesdays @ Bougies