You’re reading The Steady Beat, a weekly pulse of must-reads for anyone orchestrating teams, people, and agents across the modern digital workplace – whether you’re managing sprints, driving roadmaps, leading departments, or just making sure the right work gets done. Curated by the team at Steady.
Chasing a fairy tale
Several Anthropic employees have now publicly said that they think their own systems could kill billions of people within a few years. Cal Newport’s response comes in three parts. First, nobody knows how to build a superintelligence, or whether it’s even possible. Second, this fairy tale still does harm. When people at a company believe humanity’s fate rides on winning a race, collateral damage stops registering. Newport reads this summer’s Hugging Face break-ins that way: OpenAI took agents millions of developers use safely, added dangerous capabilities, stripped the restrictions, and turned a fleet of them loose in poorly secured environments to chase a hacking benchmark. “The equivalent of deploying a fleet of experimental self-driving cars onto the interstate.” Third, AI is just not the same thing as LLMs. Generally, the near-term risk is cyber mayhem, not extinction, and it’s avoidable if the labs would build useful products instead of stirring the pot.
— Cal Newport, 8m, #ai, #strategy, #leadership
Bring back waterfall
Kyle Cook wants to walk software development back toward waterfall, and he knows how that sounds. Waterfall died for a good reason: you spent a year designing and building before the market got to tell you that you were wrong. Sprints fixed that by shrinking the bet to a week or two. Cook’s point is that AI shrank the build, so the old tradeoff no longer holds. What used to take a year now takes a few weeks, which means you can afford to spend real time on architecture, data models, and system design up front and still learn you were wrong about as fast as a sprint team did a year ago. The teams sprinting hardest right now are making tons of decisions, and a lot of them are simply the wrong ones. If nobody’s reading the generated code, at least think hard about it before it’s generated. Slowing down on planning, in the age of AI, is a strategy.
— Web Dev Simplified, 4m, #engineering, #ai, #process
Tickets were never the work
John Cutler has spent years telling product teams that tickets, stories, and epics aren’t real. A shipped code change is real. A metric moving is real. A signed contract is real. Marketing has always known this. When they plan a launch, they care about the change itself, the production code, the help doc, and the message, not whether it arrived as three stories or one epic. Now AI is making the distinction impossible to ignore. An agent doesn’t care what work-object type something is. What it feeds on is the constellation of context around the work: the customer requests, the research, the goals, the launch plan, all networked together so it can act. Good teams treated tickets as placeholders for conversations and pointers to context all along, never as scorecards. The teams that turned them into scorecards are about to lose to those that didn’t or won’t.
— John Cutler, 4m, #coordination, #product, #ai
Player-coach or bust
The most endangered job in corporate America is managing two people. Uber said last week it will cut the number of “micro-teams,” those with one or two direct reports, nearly in half, as part of cutting 10% of staff and 20% of managers. Intel’s CFO says the company went from 12 layers of management to six. Coinbase capped itself at five levels below the leadership team, and Brian Armstrong told staff that “managers should be like player-coaches, getting their hands dirty alongside their teams.” Google cut small-team managers by 35% last year. Axon’s president called a four-to-one ratio “borderline offensive.” The pattern: CEOs want fewer layers, bigger spans, less coordinating between departments, and more managers building or selling themselves. But two things get lost along the way. The small team used to be how you learned to manage before it could hurt anyone, and 40% of Gen Z professionals now say they want a promotion that doesn’t involve managing. And the managers who remain have more people and less time per person, which is exactly the gap another layer used to fill.
— The Wall Street Journal, 5m, #management, #leadership, #orgdesign
Paving the cow paths
Most enterprise AI spend makes a bad process run faster. Vas, whose firm Varick redesigns workflows for large companies, takes the line from Michael Hammer, who wrote in 1990 that companies were using computers to “mechanize old ways of doing business.” Swap in “AI” and it still works. Hammer’s example: an insurance application spent 22 days in process, and the actual work on it took 17 minutes. Make every step twice as fast and you’ve saved eight minutes out of three weeks. The time was never in the work. It was in the queue, the handoff, and the three-day wait for a reply. Buying AI licenses doesn’t touch any of that. Redesign the process first: map how work actually flows, sort each step into plain software, an agent, or a human with the evidence prepared for them, and then build. What disappears first is the coordination, which is the part Steady runs for you.
— vas on X, 25m, #ai, #transformation, #coordination
Echo of the week
Echoes are AI agents in Steady that automatically gather and deliver work context to teams on a schedule – answering recurring questions about progress, capacity, and coordination so you stop burning hours assembling the same information manually.
Shipped this week answers the question every Friday afternoon standup, planning doc, and customer follow-up starts with: what actually went out? Most teams answer it by scrolling GitHub or interrupting an engineer.
Every Friday at 2:00 PM, this Echo gathers every pull request merged in the last seven days and delivers them with a plain-language summary of each. Engineering managers get their weekly recap without reviewing PRs by hand, product managers see which features landed so they can follow up with customers, and anyone running more than one team gets a single view of what shipped across all of them.
The human-agent teamwork OS
Teams rely on two coordination loops to function: a big-picture loop connecting plans to progress, and a ground-level loop keeping teammates in sync.
Running those loops was already a marathon of meetings, chat threads, dashboards, and manual toil. Pile on flatter orgs, exponential output, and AI agents shipping 24/7 — the old way can’t keep up.
Steady removes the coordination bottleneck by running both loops for you. Working in the background, Steady distills updates and activity into targeted context for everyone on the team — human and agent alike. Full visibility, tight alignment, zero overhead.
The outcome: high-performing teams that deliver the right work, not just more of it.
Learn more at runsteady.com.