Quinn

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

$1,612

141 sessions · $11 each

Savings opportunity

$965

60% of spend, at list price

Practices on target

3 of 7

Focus: Cache hit rate, Opus on routine work

Cache hit rate

66%

Team 78% · org 82%

FY2027 to date · estimated at list price

$965 of $1,612 spend

Largest saving first

Improve prompt caching$536
Cache hit rate 66% 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$351
82% of routine spend ran on Opus: code research $501 of $605, documentation $204 of $249, ops $26 of $32, presentation $106 of $130.
Run docs, decks, ops and read-only research on Sonnet instead of Opus.
Keep context lean$56
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$22
65% of 1,296 Bash calls ran through rtk; 1.8M 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 Data & Analytics and the organization

Practice	Quinn	Team	Org	Target	Recommendation
rtk coverage	65%Needs attention	60%	56%	≥ 80% of Bash calls	Install the rtk hook so test, build, git and search commands return compact output.
Knowledge MCP use 5 findings saved back	31%Needs attention	30%	30%	≥ 40% of sessions	Search the knowledge MCP before re-researching a codebase, and save findings back with remember.
Cache hit rate	66%Off target	78%	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	8%	10%	≤ 10% of input	On target
Opus on routine work	82%Off target	55%	44%	≤ 25% of routine spend	Default routine sessions to Sonnet and keep Opus for harness, design and hard debugging work.
Sessions using subagents	49%On target	45%	50%	≥ 30% of sessions	On target
Tool error rate	2.9%On target	3.4%	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	–	–	45%Off target	6%On target	88%Off target	26%Needs attention	$20
Mar 2026	–	–	52%Off target	10%On target	88%Off target	31%On target	$39
Apr 2026	–	–	51%Off target	9%On target	88%Off target	35%On target	$35
May 2026	29%Off target	–	53%Off target	10%On target	86%Off target	39%On target	$84
Jun 2026	52%Needs attention	17%Needs attention	61%Off target	10%On target	86%Off target	45%On target	$181
Jul 2026	76%Needs attention	23%Needs attention	63%Off target	11%Needs attention	82%Off target	40%On target	$137
Aug 2026	70%Needs attention	36%Needs attention	68%Off target	10%On target	81%Off target	55%On target	$226
Sep 2026	64%Needs attention	31%Needs attention	71%Needs attention	7%On target	79%Off target	61%On target	$71
Oct 2026	72%Needs attention	39%Needs attention	71%Needs attention	7%On target	79%Off target	53%On target	$172

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	2,314	30.7M	92.6M	7.9M	3.4M	$249
Opus 5	3,412	49.4M	129.5M	11.4M	4.6M	$499
Opus 4.8	2,114	40.2M	78.2M	7.6M	3.3M	$369
Opus 4.6	1,270	30.2M	39.7M	4.5M	1.9M	$245
Opus 4.5	38	1.2M	1.1M	144K	66K	$9
Sonnet 5.5	709	10.2M	30.7M	2.6M	1.1M	$41
Sonnet 5	1,198	17.4M	43.9M	3.9M	1.5M	$103
Sonnet 4.6	415	10M	16.3M	1.7M	714K	$52
Sonnet 4.5	9	330K	306K	41K	14K	$1
Haiku 5.5	1,486	22.6M	68.2M	5.8M	2.6M	$5
Haiku 4.5	978	16.4M	35.6M	3.3M	1.4M	$31
GPT Realtime (meetings)	56	1.1M	2.1M	207K	100K	$8
Total	13,999	229.8M	538.3M	49M	20.7M	$1,612

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.