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.
Tokenmaxxing EOL
Earlier this year the order from on high was simple: use as much AI as humanly possible. Employees called it “tokenmaxxing,” Meta and Amazon ran leaderboards, and burning the most tokens became a proxy for being AI-savvy. Then the invoices landed. Meta now caps usage after an “exponential” cost spike, Uber torched its entire 2026 AI budget by April, and the leaderboards have quietly disappeared. Welcome to “tokenminning.” The whiplash – months, not years – tells you nobody has this figured out yet. The root error was a lazy metric: leadership couldn’t measure who used AI well, so they rewarded who used it most, and volume steamrolled efficiency. Proposed fixes include reserving frontier models for the genuinely hard tasks and routing everything else to cheaper ones; AT&T says that swap alone saves up to 90%. Pick metrics that map to value, not the ones that are easy to count.
— The New York Times, 6m, #ai, #leadership, #productivity
Cost enters the chat
“We created a monster,” a Workato exec said after Anthropic flipped them to token-based pricing and their spend jumped 7x overnight. He’s not alone. Amazon, Walmart, Cisco, and Uber are all capping usage, nudging people toward cheaper models, or asking whether each automated task is actually worth it. The structural shift: agents replaced chatbots, and an agent burns far more compute – Cisco figures 10 to 1,000(!) agents per human, all grinding away around the clock. Flat subscriptions hid that; token billing exposes every prompt to the CFO. Sam Altman concedes cost became a “huge issue” this year that simply didn’t exist last year. Goldman projects a 24x jump in token consumption by 2030, so this pressure isn’t a blip. There’s a cultural move, too: Workato swapped its twice-weekly “AI enablement” sessions for “AI financial responsibility.” If your team adopted agents assuming compute was basically free, watch out. It turns out the bill scales with your ambition.
— Financial Times, 7m, #ai, #strategy, #management
The new tech debt
A room full of CTOs in Toronto arrived at the same conclusion: cognitive debt is the new technical debt. “AI makes it cheap to write code,” one put it. “That’s not the same as it being cheap to ship it, or maintain it.” That gap is the whole problem. When generation is nearly free, teams skip the quality gates that scarcity used to force on them, like build-versus-buy calls, maintenance planning, and answering the basic question of whether this code should exist at all. Throwaway internal tools calcify into permanent infrastructure nobody understands. The constraint has moved off typing speed and onto human judgment, plus an org’s ability to absorb everything the machines produce. One CTO rebuilt his entire interview around code review, “because that’s what we’re actually doing now.” Another named the real bottleneck – generation outpaces anyone’s capacity to review it, so PRs pile up unread and unowned.
— ShiftMag, 8m, #ai, #engineering, #leadership
Right work, first try
Every story above circles the same root cost: agents that move fast without context rack up debt, in dollars and in rework. Fewer agents won’t fix that. Better-informed ones will, and that’s the thinking behind two new things from Steady this week – a standalone CLI and a set of agent skills. The CLI brings Steady into the terminal: check your digest, surface blockers, post a progress update without leaving the shell, with JSON output so scripts and agents can read it too. The skills library teaches an agent to actually use it — how to operate the CLI, build against the API, write a real check-in, and, the important one, pull live team context before it acts. That last piece is the antidote to the cost problem. An agent that knows what the team is doing and where the goals sit doesn’t burn tokens chasing the wrong thing; it ships the right work the first time. Redundant effort at agent speed is the most expensive kind there is.
— Steady, 4m, #ai, #coordination, #agents
Keep the struggle
Nuno Lopes articulates what every senior engineer silently laments these days: AI removes the friction that built you. Struggle is the mechanism that turns information into intuition, and an LLM that hands you the answer quietly skips it. He points to the 2025 MIT study where heavy LLM users showed the weakest brain connectivity and couldn’t recall work they’d just produced — cognitive debt (there it is again), now at the individual level. Ship code you can’t explain often enough and you hollow out the judgment your whole career rests on. His remedy: don’t refuse the tools, just spend the hours they free up on harder thinking instead of less of it. Ask “why” after every answer, and type out the snippet instead of pasting it. Read the foundational texts. Build the skills no model owns, like debugging across systems, reading organizational context, and talking to the humans who set the goals. The time AI gives back is only a gift if you reinvest it in thinking.
— Nuno Lopes, 10m, #engineering, #ai, #leadership
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.
The Morning Paper answers a question every team lead quietly has: how do you get anyone to actually read the Monday recap? It takes last week’s progress and lays it out as a newspaper front page — a headline story on the biggest win, supporting updates beneath it, a read on the team’s mood, and a classified ad for the one task still sitting unclaimed. Same information you’d otherwise bury in a status doc nobody opens, except this one people finish. It runs every Monday at 8 AM, so the week starts with shared context instead of a scramble. Make the recap something the team looks forward to, not another thing they skip.
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.