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.
Agents get a seat
How do you tame agents gone wild? First, make them listen to the whole team. Second, make them tell you the future. In Steady, autonomous AI agents can now join an account as full members and work inside the same coordination loops as people: writing async check-ins and working off of cross-team goals. With check-ins, agents will not only tell the team what work they already did, they’ll also declare what they are going to do and where they are blocked. So humans (or even other agents) can get them back on track before the wrong work (and token spend) happens. Not only that, but the agents will pull current goals and other team check-ins into their context window before they do any work. So if priorities shift, or someone else (or another agent) picks up a chunk of work, you don’t end up with a giant mess of useless PRs alongside a huge token bill. Steady is agent-host agnostic: a simple skill and token is all your agent needs to be on the team, so it works with the *claws, Hermes, LangChain, Crew, custom code, and more. But if you’re looking for a super simple and durable way to deploy agents, we open-sourced the one we built for ourselves: a framework called OpenRoutines.
— Steady, 4m, #agents, #coordination, #product
Founder mode’s expiration date
Paul Graham’s famous “founder mode” was built for small teams. At 10 people you can run a hive mind, and even at 30 a founder making every call is still leadership. Any bigger than that and it turns into a bottleneck. Brian Elliott watches it spread anyway, as company leaders under pressure to get AI right read Paul Graham’s essay as permission to micromanage everything. The founder mode essay originally grew out of Brian Chesky centralizing product planning when Airbnb was fighting for its life during COVID, and it worked – as an emergency measure. Kelly Monahan writes, in her upcoming book Reclaim the Plot, that it outlived its emergency and got rebranded as a philosophy in the years since. The AI moment calls for the opposite: decisions pushed down to where the information sits – on the front lines, not at the top. As Monahan puts it, “AI is becoming a very expensive way to realize how poorly designed our workplaces were in the first place.”
— Work Forward, 11m, #leadership, #organization, #ai
A team, or a collection of individuals?
A year or two ago, James Socol’s team was tight-knit and high-performing: one or two shared goals a sprint, work broken down so people could run in parallel, someone always ready to jump in when a teammate got stuck. Now, thanks to AI, they ship more than ever – and with as many goals a sprint as engineers. Each person ships complete work on their own instead of splitting up features, and team meetings became status readouts the rest of the room half-listens to, since nothing in them affects anyone else. The research on what makes a team work points to shared goals where the group succeeds or fails together, interdependent work, and stable membership. AI erodes the first two. Solo shipping means fewer reviews and fewer pairing sessions, and those were the moments where culture, “the way we do things around here,” got passed along with the battle scars. Socol’s argument is that staff+ engineers have to rebuild culture on purpose: celebrate the choices you want more of, write the team charter and actually refer to it, and model the behavior out loud. It’s the remote-work shift all over again.
— LeadDev, 5m, #engineering, #culture, #leadership
What not to build
Hiten Shah looks at Meta – Stories everywhere, then Reels, then Threads – and sees something more useful than shamelessness: a company that treats the rest of the consumer internet as an external product lab, letting other people spend years and millions discovering which behaviors stick. Every software company is about to need that muscle. Implementation used to be the scarcity that forced prioritization on you: ten good-sounding ideas, capacity for two. Agents are erasing that protection. “Can we build it?” gets boring; “should we?” gets harder. We’re about to get extremely good at building bad ideas quickly, he warns. The discipline he proposes: don’t clone features, find the mechanic underneath. Look under a pull request and you find work made visible, feedback attached to it, asynchronous inspection, a state change on approval, and a history that sticks around – machinery that travels to marketing, finance, legal, and AI-generated work. The trick in Stories was lowering the pressure to create. Decide which behavior matters for your product first, then go hunting for proven machinery that creates more of it. As building gets cheap, judgment, taste, and knowing when to leave the product alone become the job.
— Hiten Shah, 9m, #product, #strategy, #ai
Know your tics
Remember Manager Readmes? The 2018 fad where a manager wrote up their leadership style and principles, and the employees handed one were rightly furious, because it reads like a manifesto. Rands wrote his anyway, and says the point was never to share it; it was to write it, to force himself to think through what mattered to him as a leader. Eight years on he’s back with a companion list: fifteen irrational, even embarrassing tics, written as if-then statements. If you tell him what to do, the hair on the back of his neck stands up; but ask for help with the same task, no problem. If he hands you a task, he has already re-graded you on every prior task, and that grade decides how much he hovers. If you give him critical feedback, he rages first, rationalizes second, and only then – after years of practice – thinks. Pick the leader you admire most: for every attribute you admire, there’s an equal amount of strange-to-broken behavior tics you never saw. The competent ones know their tics and have built the compensation techniques that keep them away from the people around them.
— Rands in Repose, 4m, #leadership, #management, #self-awareness
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.
Intentions vs. reality asks the question every Friday at 2 PM: did the week go as planned? It compares what the team said they intended to do against what they actually reported doing, then delivers the follow-through rate, the intentions that carried over unfinished, and the significant work that got done but was never stated as a plan. Socol’s team had as many goals a sprint as engineers, and Rands had an unspoken plan in his head the team kept drifting from. This is how the drift shows up in the open, with agents’ work in the comparison too, before it turns into a list of boxes that must be checked.
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.