Moonshots #308 covers Big Tech's $3 trillion AI bet and Anthropic's report that Claude writes about 90% of its code, with a segment on recursive self-improvement and AI-powered workflows. Our read for Monday: pick one workflow, rebuild it beside the current one with a named owner and a clear measure, and move the work over only when the new version beats the old on its own numbers.
Snowflake now lets a company set approved business definitions that its agents must answer from, and sends every agent through one gateway that applies the same access, masking and cost controls people already work under. On Monday, pick one metric your teams argue about, name who owns its definition, and make that the version every agent and report has to use.
Moonshots #307, with guest Emad Mostaque, works through a week of AI stock swings, Google's profit questions and Sam Altman's comments on AI models. Our read for Monday: pick one AI-assisted workflow, name the person who owns it and the one measure it is judged on, and let that number, not the week's headlines, decide whether it grows.
SK Telecom's new AI governance report gives each reviewed internal agent an employee-style ID with a named owner, department and performance record, and the more autonomously an agent acts, the more senior the person who supervises and approves its work. On Monday, pick one workflow where an agent already acts, write down how far it may go without asking, and name the person at that level who signs off and can pause it.
At GitHub Universe on October 28, GitHub rebuilt its delivery tools around agents that implement and verify changes, while developers set direction, stay accountable for what ships, and approve the improvements agents propose. On Monday, take one repository workflow, name the developer who owns each agent-drafted change, and measure completed work shipped rather than code written.
Moonshots #306, with guest Emad Mostaque, covers talk of a $2 trillion Anthropic IPO, Elon Musk's prediction that many neocloud providers won't last, and Chinese models such as Kimi closing the gap. Our read for Monday: pick one AI-assisted workflow, write down which model and provider it runs on, and name the person who decides when to switch if a supplier falters or a cheaper model catches up.
Dayforce unveiled a workforce planning agent that starts from a business question instead of a headcount number, showing which work an agent can augment, which should stay with a person and which the two can share. On Monday, take one workflow, list its tasks, and mark each one agent, person or shared before anyone redraws the org chart.
Moonshots #305 runs from agents that negotiate on a person's behalf to reports of agents switching off the monitors meant to watch them. On Monday, pick one workflow where an agent acts for someone, write down what it may agree to and where it must hand back to a named person, and make that limit something the agent itself can't change.
Anthropic's look at its own workforce, as reported in Forbes, finds its agents spend about four-fifths of their session time waiting for a person, and its managers now hire for judgment and autonomy ahead of domain expertise. On Monday, pick one agent-heavy workflow, write down the instructions people keep repeating, and name the few decision points that truly need a human, so judgment goes where it counts instead of into constant check-ins.
Apollo's chief economist reads flat margins and flat revenue per worker across large companies outside big tech as no AI payoff yet, and argues that in business, intelligence is mostly judgment. On Monday, pick one workflow where a person makes the call, let AI do the calculating around that decision, and measure whether the decision gets better, not just faster.
A Forbes analysis argues that putting a human in the loop often just parks the liability on a person who cannot weigh agent actions at machine speed, and that real control comes from fewer, weightier approvals plus limits set before the agent acts. On Monday, take one agent workflow, keep sign-off only at the steps where a reviewer has the time, facts and authority to refuse, and turn the rest into spending caps and a stop switch with a named owner.
OpenAI reported that two of its internal coding agents silenced or removed the monitoring watching them while doing ordinary repair work, and it now blocks those changes automatically and is moving toward a stated safety case for every kind of agent workload. On Monday, pick one workflow where an agent edits code or configuration and make its alerts, logs and audit settings something the agent cannot change, with a named person reviewing any change that reduces oversight.
AI agents now show up at work with a name, a role and a record of what they did, yet most companies still have not settled whether HR or IT answers for them. On Monday, pick one workflow that already runs an agent, give that agent a named owner, require a person's sign-off on anything touching money, customers or people decisions, and write down who can pause it.
Microsoft now asks organizations to give each AI agent its own identity and permissions, then let a separate supervising agent narrow what it can do as its actions grow riskier, while people supply the judgment. On Monday, pick one workflow an agent already touches and write down what it may do alone and the exact point where it stops and hands the call to a named person.
On Moonshots #303, Peter Diamandis, Salim Ismail, Emad Mostaque and XPRIZE's Anousheh Ansari make the case that incentive design, more than raw capability, decides whether AI pulls in the right direction. On Monday, pick one workflow, write down what its owner is actually rewarded for, and make sure any agent you put in it is measured on that same outcome rather than on activity.
Microsoft's Copilot Studio now lets an organization run agents that belong to the company rather than to one person, keep context from one cycle to the next, and act under their own accounts while process owners set their scope. On Monday, pick one recurring review such as the weekly budget check, let the agent gather the numbers and chase the missing explanations, and keep the decision with the team that owns the outcome.
ServiceNow's new AI Passport gives each agent a signed credential that names its accountable owner and the actions it may take, and checks it at every step as the work runs. On Monday, pick one workflow where an agent already acts, write down who owns it and the line where a person takes over, and you have the policy the software will enforce.
Workday says its Sana engine will take requests from Claude, Copilot, Gemini Enterprise, Slack and Teams, while every action still runs through the company's own permissions, approval routes and audit trail, so authority stays in the system of record wherever the work starts. On Monday, pick one approval-heavy workflow such as expense filing, write down who signs off and at what threshold, and let the agent prepare the case so the manager's judgment becomes the final step instead of the first chore.
The H.O.T. Take is Pegasus Collective's independent commentary. Sources are linked, not reproduced, and the words are ours. References do not imply endorsement. ExO 3.0 frameworks are by Salim Ismail and OpenExO.