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.
Proof of thought
Steve Huynh spent 18 years at Amazon, where Bezos had banned PowerPoint and meetings opened with thirty minutes of “study hall,” everyone reading a six-page brief in silence. If someone handed him six pages, “I knew that you had spent significant time with the problem.” AI broke that signal. Huynh describes a senior product manager presenting a document when “it was blindingly obvious to everybody in the room that he was reading the document for the first time.” Nobody said anything because he was the most senior person there. Huynh’s proposed fix is to borrow a policy from Clay’s Sophie Alpert: you own every word, generated or typed, and if the room can tell the AI did the thinking, “there must be consequences.” Shareholder letters, planning docs, PR-FAQs, and promotion documents stay human-written. The six pages are meant to be a receipt for the thinking behind them.
— A Life Engineered, 8m, #ai, #leadership, #culture
RoboGym
Imagine a gym where you press a button labeled “Lift” and watch a weight rise and fall. “Not the point of a workout? Correct.” Kent Beck calls it RoboGym, and “genies [LLMs] are being designed as a RoboGym for the mind.” Another way to frame the problem: there is a huge difference between toil and struggle. Both are work that has to get done. “The difference is in the consequences when the work is done.” Beck recently asked a genie to reformat a 2013 essay. “Spit, spot, done.” Doing it by hand would have left him “the same person. No satisfaction. No learning.” Then the other kind: three hours on a train as a college freshman, stuck on a discrete math theorem he knew was true and couldn’t prove. When he got it, he’d learned something, and he had “confidence that I could tackle problems far harder.”
— Kent Beck, 3m, #ai, #engineering, #learning
Back in the code
Bjorn Roche manages engineers, and AI “has changed my day-to-day less than the folks I manage,” but it still has changed his job. When his team “recently released a broken feature that took much too long to revert,” the missing piece was feature flags. Roche built a simple system himself in a few hours with an agent, and the team now uses it for critical deployments. Hands-on work, he says, “helps me build trust with the team” and keeps him close to systems he would otherwise only hear about. Learning an unfamiliar corner of the codebase is now a conversation with Claude, helped by READMEs and CLAUDE.md files the team has been “organically improving.” He is less resistant to a new language or service, because the ramp is shorter. The bus factor scares him less too: “Having only one or two people who deeply understand a critical system used to be a major concern.” The tools can tell you what the code does, not what the docs say it does. His caution is reserved for executives who “view AI as a magic wand for getting work done” with no vision for what to build. “The best ideas will come from people using the tools every day.”
— Bjorn Roche, 3m, #ai, #management, #engineering
Spend meets ship
Gergely Orosz collected AI spend from 27 companies and asked whether any of it shows up as shipped software. A quant trading shop with about 2,000 developers went from $500K a month in January to $8M in September, roughly $1,600 per developer. A two-person .NET agency pays $3,000 per developer per month and went from one or two PRs a day to more than ten. In CodeRabbit’s data, weekly PRs per developer are up 113% year over year. However, the outcomes are mixed: a 600-engineer SaaS company sees more PRs and about the same merge velocity. A 5,000-person public company has flat PR frequency and PRs three times the size. A 100-person AI company with a $2,000 allowance reports velocity “unchanged.” One engineer: “I personally see accelerated slop produced all around me.”
— The Pragmatic Engineer, 12m, #ai, #engineering, #metrics
Green isn’t proof
Leonid Bugaev spent time with factory workers, and “they told me they do not trust software engineers.” They’d watched engineers’ apps fail on the floor: dust on a sensor, parts from a different batch, a guard somebody bypassed. He thinks they have a point. Most software “engineering” is a pipeline. Ticket, code, tests, pull request, “then the pipeline turns green.” Requirements live in the ticket and the rationale lives in someone’s head. “A green pipeline tells us that the checks we selected passed,” which says nothing about whether they were the right checks. Real engineering disciplines start from stakeholder needs and hazards, treat coding as one step, and separate verification (built right) from validation (the right thing for its environment). “Code and tests matter, but they are replaceable artifacts.” The requirement and its evidence should outlast them. Bugaev’s bet: “AI may not be the end of software engineering. It may be the first thing forcing us to finally practise it.”
— The Verification Gap, 9m, #engineering, #quality, #ai
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.
Progress report. The weekly status update exists because a manager wants to know what each person worked on, and the only way to find out was to ask. This Echo answers without asking. Every Monday at 11:00 AM, it summarizes what each person on the team worked on over the past seven days, drawing on check-ins and activity from connected tools. Managers walk into the 1:1 already briefed, skip-levels see individual work without routing through the direct manager, and nobody writes a status report on Friday afternoon.
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.