// ACTIVE LOOPS: 14 · // DATA ROUTED (24H): 1.2TB · // LAST DEPLOY: 14 MINS AGO · // MCP GATEWAYS ONLINE: 47 · // LEGACY ERP BRIDGES: 23 · // BACNET NODES: 1,284 · // UPTIME (30D): 99.987% · // QUEUE DEPTH: 3
Playbook2026-05-03· 7 min read

Why we cap at one active request

Concurrency is not productivity. The arithmetic of context-switching when AI is in your toolchain.

Concurrency simulatorTeam capacity estimator

When a senior engineer juggles three projects, each one moves at one-third speed plus a context-switching tax. With AI in the loop, the tax is steeper — every switch costs you a freshly rebuilt context window.

So Axiom Core caps at one active request. Scale at two. Requests close in days, not quarters, and your problem gets the team's full attention for the short window it actually needs.

// Concurrency simulator

Move the sliders. The "throughput per project" line is the part teams underestimate the most.

interactive
live · stateful

N concurrent projects, one team

Assumes a 5-day-per-project base velocity. Tax compounds per switch.

Active projects3
Context-switch tax20%

Days to close one

7.0d

Team throughput

0.7/wk

Projects shipped

Effective velocity

71%

Of theoretical max

Project 120% done
Project 219% done
Project 319% done

When N=1, efficiency is 100% by definition. By N=4, even a modest 20% tax cuts effective velocity below half. That's the whole argument for the cap.

The math isn't subtle. The pushback is always cultural: but we'll feel slower. You won't. You'll just see fewer half-finished branches sitting open on Friday.

// Your team's real cap

Plug in your team size and average request shape. We'll tell you how many active requests you can carry before context-switching eats the gains.

interactive
live · stateful

Capacity estimator

Most teams find their honest cap is one or two below what they currently run.

Team size3
Request complexity50%

Honest cap

2

Concurrent active requests

Verdict

Boring

Calmer Mondays

Two at most. Three needs a structural change.

// SIGNAL BACK · one-request-at-a-time

one tap

// NEXT

The boring stack we ship on in 2026

Read next