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.
Code Outpaces Trust
The always insightful Kent Beck just named the problem every team using coding agents is facing: accumulating code faster than accumulating trust. That gap is a big problem, because code is only an asset if people trust it enough to build on it without re-checking everything. And the two grow at different speeds: code piles up in an afternoon, while trust builds slowly and evaporates instantly. And once a teammate stops believing in your output, every line you generate with an agent isn’t velocity – it’s a backlog of work someone else has to verify by hand. Beck reaches back to Extreme Programming for the solution, which he reframes as a “trust factory”: practices like pairing, continuous integration, and customer involvement weren’t really about code quality. They were trust-building machinery: each practice makes the work more reliable and, just as important, makes that reliability visible to the people depending on it. Developing blindly with coding agents dismantles that machinery. An agent optimized to satisfy your prompt isn’t optimizing for correctness, maintainability, or the colleague who inherits the mess – it’s optimizing for speed. So slow down. More human interaction, more structural improvement, more alignment on purpose. As Beck and XP taught us before, the teams that invest in trust ship faster, because trust is the thing that lets you actually use all that velocity.
— Tidy First?, 5m, #ai, #engineering, #trust
The Best Riders Win
Tomasz Tunguz has a two-minute read that explains why your AI pilot stalled: the model was never the product. “Like a mustang, AI is powerful but wild. Harnessing the power means domestication.” The raw intelligence everyone’s racing to adopt only becomes useful work inside a harness – and Tunguz maps the seven components every serious one needs. Context and memory tuned to your domain. Tools exposed safely, with validation and approval gates. Orchestration loops that think, act, and observe. State that survives failures. Sandboxed compute. Observability so you can trace what the agent actually did. And cost discipline about which model handles which job. Notice what that list really is: a management system. Onboarding context, scoped permissions, checkpoints, oversight, escalation paths – it’s the same infrastructure you’d build around a talented new hire who moves fast and lacks judgment. The advantage goes to whoever builds the better harness – in software, and in how their teams work.
Editor’s note: way easier said than done. Continuous Coordination, anyone?
— Tomasz Tunguz, 2m, #ai, #systems, #strategy
Function vs. Relationship
Gaurav Jain’s answer to the “will AI replace managers?” debate rests on a simple distinction: management is a function, leadership is a relationship. A function can be optimized. A relationship has to be earned. The administrative load that fills a manager’s week – performance documentation, status summaries, stakeholder updates, chasing information across tools – is function work, and AI is already absorbing it. Good riddance. But Jain’s point isn’t that managers are obsolete; it’s that the job is finally getting clarified. “AI can’t sit across from Sarah on your team who is burning out, and know that what she actually needed is not a ‘performance plan.’” No model coaches someone through a career crisis, makes the unpopular call, or earns trust over months of consistent behavior. Value flows to what’s scarce, and as machines absorb the administrative layer, human judgment and genuine connection become the scarce assets. Which means the excuses are gone too. So when the busywork disappears, the only question left is whether you’ll use the recovered hours to actually lead.
— The Good Boss, 6m, #ai, #leadership, #management
The Glue Trap
Tanya Reilly’s “Being Glue” is timeless, but now’s the time to re-read it. The setup: a mid-level engineer spends two years doing the work that actually makes her project succeed – onboarding juniors, fixing communication breakdowns, writing things down, unblocking everyone. Glowing reviews. Then the promotion committee shrugs: “You’re not producing much code.You didn’t have enough impact yet.” (In today’s world, maybe it’s “not using enough AI tokens.”) The project shipped because of her, and the system can’t see it. Worse, the burden isn’t distributed evenly: women volunteer for non-promotable work 48% more often than men, and managers hand it to them 44% more often. That’s a machine for pushing your best coordinators out of engineering. Reilly’s advice: if you only do glue, you only get better at glue, so make explicit deals with your manager about what gets rewarded. And if you run a team: glue work is real work, especially with all the AI agents muddying the signal. Allocate it deliberately, name it publicly, and promote for it – or keep quietly taxing the people holding your teams together.
— No Idea Blog, 13m, #leadership, #coordination, #management
API Demos
A note from our own shop: with the latest release of our API, we took a few moments to wrap some of our projects that use into the Steady Playground, a public repo of MIT-licensed demo apps. Each one pulls work context out of Steady and puts it somewhere a team actually looks. A Docker daemon that summarizes Claude Code activity. A macOS menu bar announcer for your daily digest. A desktop display for the team’s intentions. And two apps for the Tidbyt LED display, rendering check-ins and goal progress as glanceable pixels – 64 wide, 32 tall, sitting on your desk like a tiny mission control. None of these are products. That’s the point. The web app is one way into Steady, but the API and MCP server mean your team’s coordination signal can show up in a terminal, a menu bar, or a toy on your desk. Grab a personal access token, fork a demo, and wire your team’s pulse into whatever surface you stare at all day.
— Steady, 4m, #coordination, #engineering, #tools
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.
Team Capacity. Discovering someone’s on vacation the day before a deadline is avoidable. This Echo compiles upcoming absences into a consolidated view organized by week for the next 30 days, so you can plan work around confirmed availability instead of getting blindsided by out-of-office replies. Coverage gaps, on-call rotations, deadline math – all visible before they bite. Runs monthly on a Monday, right when you’re planning ahead.
The lightweight 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 is the lightweight teamwork OS that runs both loops for you. Working in the background, it distills updates and activity into personalized intelligence 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, 3X faster.
Learn more at runsteady.com.