Wren

Engineer · AI usage, harness practices and where spend could have gone further.

Demonstration report. This engineer and their usage are synthetic.

This page is for the engineer and their manager to find cheaper defaults and harness fixes. It is never used to rank people.

Period

Summary

FY2027 to date

AI spend

$1,929

130 sessions · $15 each

Savings opportunity

$497

26% of spend, at list price

Practices on target

3 of 7

Focus: rtk coverage, Knowledge MCP use

Cache hit rate

82%

Team 84% · org 82%

FY2027 to date · estimated at list price

$497 of $1,929 spend

Largest saving first

Improve prompt caching$252
Cache hit rate 82% against an 85% target.
Keep stable context at the front of prompts and avoid switching models mid-session, so 85% of input reads from cache.
Right-size models$173
41% of routine spend ran on Opus: code research $130 of $300, documentation $24 of $78, ops $262 of $637, presentation $4 of $14.
Run docs, decks, ops and read-only research on Sonnet instead of Opus.
Keep context lean$59
9% of input tokens were past 100K of context.
Compact or /clear between tasks instead of carrying more than 100K tokens of context.
Route commands through rtk$14
64% of 1,268 Bash calls ran through rtk; 2M tokens of command output went into context.
rtk compacts test, build, git and search output before it reaches the model (about 70% smaller).

Compared with Infrastructure and the organization

Practice	Wren	Team	Org	Target	Recommendation
rtk coverage	64%Needs attention	69%	56%	≥ 80% of Bash calls	Install the rtk hook so test, build, git and search commands return compact output.
Knowledge MCP use 4 findings saved back	28%Needs attention	37%	30%	≥ 40% of sessions	Search the knowledge MCP before re-researching a codebase, and save findings back with remember.
Cache hit rate	82%Needs attention	84%	82%	≥ 85% of input	Keep CLAUDE.md and tool lists stable, avoid mid-session model switches, and resume sessions instead of restarting.
Context past 100K	9%On target	6%	10%	≤ 10% of input	On target
Opus on routine work	41%Needs attention	33%	44%	≤ 25% of routine spend	Default routine sessions to Sonnet and keep Opus for harness, design and hard debugging work.
Sessions using subagents	47%On target	51%	50%	≥ 30% of sessions	On target
Tool error rate	3.4%On target	3.1%	3.0%	≤ 4% of tool calls	On target

Is the engineer moving toward the targets?

Month	rtk	Knowledge	Cache	Context	Model fit	Subagents	Savings opportunity
Feb 2026	–	–	61%Off target	7%On target	62%Off target	37%On target	$36
Mar 2026	–	–	67%Off target	9%On target	62%Off target	33%On target	$18
Apr 2026	–	–	67%Off target	11%Needs attention	62%Off target	26%Needs attention	$46
May 2026	22%Off target	–	73%Needs attention	11%Needs attention	55%Off target	48%On target	$44
Jun 2026	47%Needs attention	13%Off target	77%Needs attention	8%On target	53%Off target	47%On target	$95
Jul 2026	75%Needs attention	24%Needs attention	75%Needs attention	7%On target	40%Needs attention	40%On target	$82
Aug 2026	72%Needs attention	35%Needs attention	77%Needs attention	9%On target	38%Needs attention	47%On target	$108
Sep 2026	72%Needs attention	33%Needs attention	85%On target	9%On target	30%Needs attention	49%On target	$47
Oct 2026	67%Needs attention	37%Needs attention	90%On target	9%On target	27%Needs attention	63%On target	$21

FY2027 to date

By model tier, by month

Routine types (documentation, presentation, ops, code research) rarely need Opus

Cache reads are billed at a fraction of uncached input; a high share is what good caching looks like

Model	Requests	Uncached input	Cache reads	Cache writes	Output	Spend
Opus 5.5	2,415	6.8M	117.3M	7.9M	3.1M	$152
Opus 5	3,016	24.4M	153.1M	11.3M	4.6M	$385
Opus 4.8	1,894	23.2M	97.9M	7.7M	3.2M	$293
Opus 4.6	1,534	22M	62.1M	5.4M	2.2M	$229
Opus 4.5	90	1.8M	3.2M	320K	158K	$16
Sonnet 5.5	3,153	8.9M	179.9M	12M	5M	$115
Sonnet 5	6,480	50.1M	314.9M	23.3M	8.6M	$461
Sonnet 4.6	2,090	24.3M	88.4M	7.2M	2.7M	$167
Sonnet 4.5	90	1.9M	3.5M	347K	143K	$10
Haiku 5.5	4,330	8.4M	212.6M	14.1M	5.9M	$8
Haiku 4.5	3,959	30.9M	170.1M	12.8M	5.4M	$91
GPT Realtime (meetings)	26	80K	1.2M	79K	29K	$2
Total	29,077	202.8M	1404.3M	102.6M	41M	$1,929

Spend and tokens come from LiteLLM gateway logs and imported agent session history, priced at API list rates.
Savings are estimates: routine Opus work repriced at the Sonnet available at the time; uncached input moved to cache reads up to an 85% hit rate; rtk removing 70% of unfiltered command output, which stays in context for about half the session's remaining turns; and 50% of input past 100K context avoided by compacting. Levers overlap, so the total is an upper bound.
rtk and the knowledge MCP are scored from their rollout (rtk May 2026, knowledge MCP Jun 2026).
Generated Oct 20, 2026, 7:46 AM.