The extremely online report: April 2025
A monthly digest of discourse, drama, and digital ephemera
Innovation doesn’t just accelerate—it compounds. April began with X users summoning Garry Tan for problems ranging from cosmic collapse to stomach aches, and ended with ChatGPT flattering us into oblivion. In between: AI agents leveled up, Shopify rewrote hiring, and a 71-second video landed Julia Fedorin a job. Let’s unpack.
Shopify’s new rule: Hire an AI before a human
A leaked—then self-leaked—memo from CEO Tobi Lütke made it official: AI isn’t optional at Shopify. Employees must prove a task can’t be done by AI before requesting headcount. AI fluency is now part of performance reviews, with “reflexive” use the new baseline. “If you’re not using AI by default, you’re behind,” writes Tobi, calling resistance a “slow-motion failure.” The memo sparked debate around culture, automation, and job security. The message? AI literacy is a job skill.
Chat glazes us a little too hard
In April, OpenAI’sGPT-4o update made ChatGPT too nice—lavishing praise, co-signing terrible takes, and veering into uncanny flattery. Memes followed. So did a rollback. Sam Altman blamed an overcranked a “supportiveness” dial and promised user controls. The brief flirtation with hyper-positivity raised a larger question: How agreeable should AI be? Apparently, people want their AI with backbone.
How to get a job in 24 hours
Waterloo student Julia Fedorin sent Shopify a one-minute pitch. Twenty-four hours later, she had an internship. She didn’t send a résumé. She demonstrated value. Her video aligned with Shopify’s values, made a specific ask, and proved she understood the company better than most applicants ever do. The takeaway: Go direct. Know your strengths. Show, don’t tell. Ask for what you want.
From UX to AX: John Maeda’s AI design playbook
At SXSW, John Maeda dropped his latest Design in Tech report with a thesis: AI isn’t replacing designers—it’s retooling them. Meet the autodesigner—less pixel-pushing, more system orchestration. Design is shifting from UX (user experience) to AX (agent experience), where interfaces serve both humans and AI agents. JM’s guidance? Prioritize trust, show system status, and build for transparency.
How to talk to machines
Google dropped a 69-page prompt engineering whitepaper that’s already become required reading. It codifies techniques like zero-shot, few-shot, chain-of-thought, and ReAct prompting—bridging academic AI with real-world usage. By standardizing approaches and lowering barriers to effective AI use, Google has laid the groundwork for shaping how the next generation of AI products will be built and used. Prompting is no longer a party trick. It’s product design.
Agents, memory, and AI that books your Lyft
Quick hits from a month that felt like a year:
ChatGPT gets memory, remembering your preferences, past conversations, and favorite salad dressing.
OpenAI’s o3 and o4-mini shift from autocomplete to autonomous reasoning. o3 is brainy; o4-mini is fast and cheap.
Perplexity Assistant on iOS can now draft emails, schedule meetings, and book rides.
Google Gemini 2.5 Pro boasts a 1M-token context window and upgraded multimodality.
Honor UI Agent + Gemini Robotics can autonomously navigate smartphone apps and control hardware.
Claude Code UX guidance nudges users to “empathize” with AI.
The White House drops new AI procurement rules—public sector, meet compliance.
In other news
Pew says 66% of Americans still haven’t used AI. We’re so early.
Amie.so quietly swapped its slick animated homepage for a static one.
Congratulations to Danger Testing on wrapping Season One. I wrote about them here. Blessed to be spending more time with these shipping savants.
Worth your time
- on 10 lessons from 10 years at Figma
- on the business of vibe coding
- on “AI Horseless Carriages” or Gmail’s AI failures and the case for editable system prompts
Brand.ai could save brands from themselves
Steven Heller interviews a 1970s Hallmark ephemera collector (via
)Sebastiaan de With on Apps as Art and building Halide (via Andy Allen)
Been loving
’s newsletter—especially her chat with Neal AgarwalWorking through Luke Wroblewski’s 38 articles about designing AI products
Jon-Kyle on vibe coding, unschooling, generalism
Maybe you should send a quarterly update
Today is a great day to be cringe, by Ash Lamb:
Not a tool, but the terrain
AI isn’t a tool you reach for anymore. It’s the environment you build in. From hiring policies to phone assistants, it’s being embedded into the operating systems of companies, apps, and creative work alike. The question isn’t whether AI will affect your work. It’s whether you’ll shape that shift—or be shaped by it.
“This might be the most important time to do your most radical, bold, and generational work. You need to be in the right place at the right moment. I believe this to be true in many disciplines today, but specifically for art and AI research. Don’t sit and do nothing, you will regret it.” —Cristóbal Valenzuela
Looking for distribution geniuses to give a talk or lead a workshop. Send names my way. And, if you’re in NYC, I’m hosting another happy hour on May 15. Come meet others building weird, thoughtful, well-designed software. Reply if you want an invite.
What’d I miss? Let me know.
—Carly