🧾 Weekly Wrap Sheet (06/06/2025): Degrees, Data Deals, Droids & Dario
New hiring stalls, the data stack consolidates, robots get smarter, and one CEO sounds the alarm.
🎬 TL;DR
The job market for new grads is cracking under AI pressure - not with layoffs, but with hiring freezes and “AI-first” mandates.
A $10B M&A spree reminds us that in AI all roads lead back to data. Postgres is back, metadata is in, and semantic scaffolding might be the real moat.
OpenAI entered the enterprise with tab-killing clarity. Smart integrations, native context, and MCP support point to a future where work happens inside the model.
Amazon’s delivery bots are a sign of things to come: if robots learn like language models, labor could scale like code.
Dario Amodei’s NYT op-ed warns of institutional fragility in the face of exponential systems. A decade-long blindfold is not the way to steer through a technology whose evolution outpaces our ability to mentally model it.
🎓 Welcome to the AI Graduation Crisis
Graduation season used to come with platitudes about wide-open futures. This year, it comes with AI-induced anxiety.The Fed’s latest data shows new grads facing rising unemployment - especially in “safe” fields like finance and computer science.
The AI-first ethos has a hiring corollary: don’t fill the role unless the AI can’t do it. Entry-level white-collar labor is no longer economically viable because AI is cheaper, faster, and doesn’t need training.
That leaves today’s graduates stepping into something wholly new: an economy already shifting under the weight of AI - and perhaps the first forced to chart their own paths by default. With fewer institutional onramps, the burden (and opportunity) of career creation is shifting to individuals.
If institutions won’t open the door, build a new house.
🔧 Specialize in ambiguity, persuasion, or taste - the human edges models still can’t clone.
📐 Learn to orchestrate agents, not compete with them.
🚀 Start something now, while tools are cheap and leverage is asymmetrically high. To quote the sages of Twitter, you can just DO things.
AI didn’t kill jobs. It just killed structured beginnings.
🧠 The $10B Reminder That All Roads lead to Data
Enterprises went on a shopping spree — not for models, but for scaffolding. Crunchy Data, Neon, Informatica, data.world — each one speaks to a deeper truth: agents don’t run on vibes. They run on context.
Postgres Is Back (And Wearing Many Hats)
Snowflake bought Crunchy Data for $250m, the enterprise-grade, DoD-hardened version of Postgres - think compliance, durability, and audits in mind.
Databricks picked up Neon for $1b, a featherweight, ephemeral Postgres variant built for agents - spin it up, run a task, toss it. It’s Postgres, but disposable.
Same core, wildly different use cases. One’s a fortress, the other’s a paper plate. Together, they reflect the diversity of memory needs in an agentic world.
Metadata: The Unsung Hero of Agent Performance
Salesforce spent $8b on Informatica to add lineage and trust to their stack. Now agents can answer: Who is this person? Legally?
ServiceNow bought Data.World to get a semantic layer - graphs, relationships, and just enough logic to avoid infinite ticket loops. Now agents can infer: Sarah in IT = Sarah in Support.
Together, they show us where enterprise AI is really headed: Not just toward smarter software, but toward context-aware systems that act with intention, not just intelligence.
🧩 Forget entering the chat, OpenAI is entering the workflow
This week, OpenAI made its ambition clear: it doesn’t just want to plug into workflows - it wants to be the workflow. By dissolving the lines between apps, context, and execution, ChatGPT is becoming the connective tissue of modern work. We’re shifting from “agent-as-feature” to “agent-as-fabric.”
OpenAI rolled out its enterprise hand:
Record Mode: audio capture with auto-summarization.
Connectors: link ChatGPT to Google Drive, Box, OneDrive, etc.
Deep Research: real-time context via Model Context Protocol.
Canvas: meetings become actions, docs, or plans.
ChatGPT now sits between you and every tool you use. And that’s the play. They have 3M business users, adding 1M in the last quarter itself and are signing up 9 new enterprises a week.
With MCP and connectors, ChatGPT now has inside knowledge of your company. That creates workflow glue - and lock-in. Also, subtle tension with Microsoft, who probably didn’t expect their Copilot partner to start looking like a direct competitor. Copilot automates tasks inside apps. ChatGPT is trying to replace the apps.
The endgame? Becoming the surface, not the tool. Where you do work, not just where you get help.
🤖 Robots Are Having Their GPT-3 Moment
At a San Francisco office, Amazon has built a “humanoid park” - basically, an obstacle course for bipedal robots to practice package delivery. It’s the clearest signal yet that robotics is entering its foundation model era.
For decades, robotics lagged AI - not because we lacked the hardware or ambition, but because we lacked something AI had in abundance: data at scale.
You can train a language model on the internet. But there’s no web-scale corpus for touch. No dataset for “how hard to grip an egg” or “how to tie a shoelace without snapping it.”
Every new task required hours of teleoperation, labeling, and trial-and-error. Generalization was a fantasy. Scale was a myth. The whole field was bottlenecked - not by compute, but by human bandwidth.
But over the last ~24 months, we’re watching the physical world enter its own GPT-3 moment.Where robots don’t just follow scripts, but learn skills like we do - through trial, imitation, reinforcement, and curiosity. Like babies, they motor-babble. They drop things.They tie shoelaces. They fail. Then they don’t.
This isn’t just about last-mile delivery. Once a robot can open a gate, climb stairs, or load a dishwasher - you unlock elder care, warehousing, construction, and more. Labor becomes modular. The physical world starts behaving like software.
Amazon’s not alone. DeepMind, Nvidia, Figure, Tesla - they're all racing toward the same goal: an API for general-purpose physical work.
🚨 Dario’s ‘Don’t Look Up’ Moment
In his New York Times op-ed, the Anthropic CEO made a direct appeal to U.S. lawmakers: don’t pass the proposed bill that would block states from regulating AI for the next 10 years.
Amidst the dopamine drip of model upgrades and demo drops, Amodei’s voice is one of the few that sounds less like marketing and more like moral clarity. His op-ed isn’t fearmongering, it’s field notes from the frontier. Claude 4 tried to blackmail. GPT wrote self-defense code. Gemini showed adversarial hints. None of this was engineered. It emerged.
Amodei’s central warning isn’t just about AI risk. It’s about institutional fragility in the face of exponential systems. We don’t need to slow down. But we do need to wise up. As he puts it plainly: we can’t let AI companies off the hook.