Avery

Lead 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

$2,277

191 sessions · $12 each

Savings opportunity

$549

24% of spend, at list price

Practices on target

3 of 7

Focus: rtk coverage, Knowledge MCP use

Cache hit rate

84%

Team 82% · org 82%

FY2027 to date · estimated at list price

$549 of $2,277 spend

Largest saving first

Improve prompt caching$352
Cache hit rate 84% 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$105
41% of routine spend ran on Opus: code research $120 of $313, documentation $109 of $245, ops $18 of $46, presentation $3 of $12.
Run docs, decks, ops and read-only research on Sonnet instead of Opus.
Keep context lean$73
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$19
65% of 1,943 Bash calls ran through rtk; 3.1M 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 Content Pipeline and the organization

Practice	Avery	Team	Org	Target	Recommendation
rtk coverage	65%Needs attention	62%	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	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	84%Needs attention	82%	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	9%	10%	≤ 10% of input	On target
Opus on routine work	41%Needs attention	57%	44%	≤ 25% of routine spend	Default routine sessions to Sonnet and keep Opus for harness, design and hard debugging work.
Sessions using subagents	57%On target	55%	50%	≥ 30% of sessions	On target
Tool error rate	2.4%On target	2.4%	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	–	–	58%Off target	6%On target	62%Off target	43%On target	$35
Mar 2026	–	–	65%Off target	11%Needs attention	62%Off target	36%On target	$122
Apr 2026	–	–	64%Off target	10%On target	–	55%On target	$41
May 2026	28%Off target	–	69%Off target	7%On target	55%Off target	42%On target	$97
Jun 2026	51%Needs attention	14%Off target	74%Needs attention	10%Needs attention	53%Off target	38%On target	$71
Jul 2026	68%Needs attention	25%Needs attention	77%Needs attention	9%On target	40%Needs attention	49%On target	$57
Aug 2026	69%Needs attention	32%Needs attention	78%Needs attention	10%On target	38%Needs attention	51%On target	$43
Sep 2026	75%Needs attention	39%Needs attention	83%Needs attention	11%Needs attention	30%Needs attention	68%On target	$51
Oct 2026	74%Needs attention	34%Needs attention	92%On target	8%On target	27%Needs attention	76%On target	$33

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	4,202	9.5M	233.8M	15.5M	6.1M	$285
Opus 5	3,127	21.1M	144.2M	10.6M	4.6M	$359
Opus 4.8	1,950	21.1M	86.4M	6.9M	3.1M	$269
Opus 4.6	3,217	56.3M	131.5M	12M	4.8M	$543
Opus 4.5	97	2.2M	3.5M	360K	146K	$18
Sonnet 5.5	6,233	9.3M	304.6M	20M	8.6M	$186
Sonnet 5	4,211	31.3M	219.3M	16M	6.7M	$320
Sonnet 4.6	2,172	30.3M	86M	7.4M	3.2M	$192
Sonnet 4.5	68	1.6M	2.6M	269K	116K	$8
Haiku 5.5	9,403	12.9M	561.7M	36.7M	13.3M	$18
Haiku 4.5	3,008	22.5M	133.4M	9.9M	4.1M	$69
GPT Realtime (meetings)	99	1.1M	4.6M	363K	157K	$10
Total	37,787	219.1M	1911.5M	136M	55M	$2,277

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.