Field guide / sub-domain
34
sources in this field
Updated September 16, 2026
Current thesis
The shortest path to orientation.
This synthesis records claims and practices from the cited sources; reported outcomes and product capabilities have not been independently verified.
Automation consolidates around parallel subagents with scoped access and hard guardrails. Scheduled Claude routines operate as 'night shift' labor with minimal setup. The approval boundary is reversibility: finish undoable work (drafting, tagging, research) autonomously; park irreversible actions (sends, spending, publishing, deletions) for human review. An n8n inbox-management agent pulls replies, classifies sentiment, checks availability, drafts responses, and pushes both to Slack for review before sending—speed-to-lead (minutes vs. hours) materially changes conversion. Setup takes roughly two minutes with a single prompt. Infra manager agents enforce hard guardrails distinct from optimization: never pause/unpause sending accounts (kills warmup) and never quarantine domains on thin data—these are treated as constants the bot never overrides. Parallel workflows require tight loops to prevent unproven changes from shipping.
Evidence board
Claims worth carrying forward01Cold outbound GTM is decomposed into four persistent GrokBot agents—list builder, outbound copywriter, campaign manager, infrastructure manager—each owning one artifact (lead list, copy, campaign build, sending health) with one owner and one definition of done.
02GrokBot gives each agent a persistent cloud computer (browser, filesystem, terminal) that logs into apps with real credentials instead of APIs, keeps running when the laptop is off, delegates to other bots in a shared thread, and can convert a recorded human workflow into a repeatable 'skill'.
03GPT-6 Astra (Sept 2026) is used as the coding layer: state of the art on computer use/browsing/software engineering, 1M token context, $10/$50 per million input/output tokens; it writes scrapers, enrichment jobs, and reporting queries behind the GrokBot agents.
04Reliable multi-agent handoffs rest on three mechanisms: naming as join key (campaign name = list name so no bot has to ask what belongs where), task-state-as-trigger (marking a task complete fires a webhook waking the next bot, no human forwarding needed), and 'blocking out loud' (a bot missing an input posts what it needs and stops rather than guessing).
05Porting existing human SOPs to agents required little rewriting because handoff points already existed where one human stopped and another picked up—suggesting well-documented human workflows are a low-friction path to agentic automation.
06A/B testing discipline for AI-generated outbound copy: enforce 'one changed variable per script' (e.g., only offer framing moves, everything else frozen) so that performance data collected two weeks later remains attributable to a single cause.
07When an agent underperforms, the root cause is usually a missing numeric threshold rather than a bad model—an unspecified judgment call gets guessed by the agent, and a repeated guess hardens into an unintended pattern.
08Infra manager agent enforces hard guardrails distinct from optimization: never pause/unpause a sending account (kills warmup) and never quarantine a domain on thin data (a few sends in the lookback window proves nothing)—these are treated as constants the bot never overrides.
Adjacent fields
Latest evidence
Levi Munneke
A cold email agency architecture uses four persistent GrokBot agents (list builder, copywriter, campaign manager, infra manager) coordinated via a shared thread, with GPT-6 Astra as the coding/data layer, to run cold email GTM as a chain of SOP-defined artifacts with strict handoff rules.
Christian
A tactical breakdown of the domains, inboxes, sequencer, list-building, messaging, and reply-handling stack needed to send 10,000 cold emails daily, with cost figures (~$1,500-1,700/month for inboxes) and a conservative conversion model projecting ~6 closed deals/month from that volume.
Charlie Hills
Anthropic's Claude Code team formally defines agentic 'loops' as cycles of work repeated until a stop condition is met, categorizing them into four types (turn-based, goal-based, time-based, proactive) by what gets handed off. Charlie Hills synthesizes the framework with practical commands (/goal, /loop, /schedule, skills) and a warning that every loop still needs an explicit stop condition or it burns tokens unattended.
Codez
A SpaceXAI engineer's tutorial on Grok Bot, a desktop app giving each agent its own cloud computer to log into tools, execute multi-tool workflows, and coordinate with other agents—shifting the user from prompting to delegating and managing approval boundaries.
Federico Viticci
Federico Viticci highlights recurring Codex-based automations on macOS he built and still uses months later, including processing voice notes, email-to-task routing, and auto-tagging read-later items, detailed in a MacStories Club guide.
Vibe Marketers HQ
A framework for solopreneurs using an agent like Grok Bot to run ten parallel marketing workflows, plus a warning that unmanaged output creates a judgment bottleneck requiring prioritization and quality gates.
The Startup Ideas Podcast (SIP) 🧃
A structured framework for using Claude Code as an 'AI employee' rather than a chatbot, covering repo-as-workspace, memory files (CLAUDE.md/roadmap.md/review.md), plan-mode briefs, scoped tickets, visual verification, three-layer review, scheduled routines, parallel agents, tiered permissions, and reusable skills/hooks — demonstrated via a med-spa missed-lead responder build.
Jason Fried
Jason Fried describes Agent Clawdito, an autonomous agent embedded in Basecamp that generates weekly/monthly customer sentiment reports, self-creates cards and to-dos, posts chat summaries, and self-documents its own methodology for human review.
Vox
A scheduled prompt run one hour before bed scans repos, TODOs, failed tests, and agent handoff notes to surface 3–5 overnight-safe tasks, but only writes a report—never edits files or starts an agent. Candidates must meet strict safety criteria and include a full paste-ready execution prompt for morning review.
Pierre-Eliott Lallemant
A LinkedIn outbound framework (used with the Gojiberry agent) combines intent signals (topics/competitor engagement) with ICP filtering, replaces cold demo asks with value-permission offers, and automates lead-finding rather than relationship-building—yielding 59%+ connection rates and 300+ demos in 30 days.