Riley

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

$3,254

184 sessions · $18 each

Savings opportunity

$966

30% of spend, at list price

Practices on target

4 of 7

Focus: rtk coverage, Cache hit rate

Cache hit rate

77%

Team 84% · org 82%

FY2027 to date · estimated at list price

$966 of $3,254 spend

Largest saving first

Improve prompt caching$664
Cache hit rate 77% 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$177
31% of routine spend ran on Opus: code research $146 of $487, documentation $130 of $383, ops $129 of $424, presentation $30 of $92.
Run docs, decks, ops and read-only research on Sonnet instead of Opus.
Keep context lean$114
10% 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$11
77% of 1,812 Bash calls ran through rtk; 2.7M 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 Agent Platform and the organization

Practice	Riley	Team	Org	Target	Recommendation
rtk coverage	77%Needs attention	74%	56%	≥ 80% of Bash calls	Install the rtk hook so test, build, git and search commands return compact output.
Knowledge MCP use 11 findings saved back	50%On target	48%	30%	≥ 40% of sessions	On target
Cache hit rate	77%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	10%On target	6%	10%	≤ 10% of input	On target
Opus on routine work	31%Needs attention	29%	44%	≤ 25% of routine spend	Default routine sessions to Sonnet and keep Opus for harness, design and hard debugging work.
Sessions using subagents	59%On target	59%	50%	≥ 30% of sessions	On target
Tool error rate	2.8%On target	2.5%	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	–	–	62%Off target	10%On target	42%Needs attention	42%On target	$127
Mar 2026	–	–	61%Off target	9%On target	42%Needs attention	29%Needs attention	$121
Apr 2026	–	–	64%Off target	9%On target	42%Needs attention	47%On target	$161
May 2026	28%Off target	–	68%Off target	9%On target	37%Needs attention	42%On target	$221
Jun 2026	55%Needs attention	24%Needs attention	75%Needs attention	10%Needs attention	36%Needs attention	50%On target	$114
Jul 2026	83%On target	38%Needs attention	75%Needs attention	7%On target	26%Needs attention	47%On target	$72
Aug 2026	90%On target	59%On target	79%Needs attention	10%Needs attention	25%On target	75%On target	$105
Sep 2026	89%On target	60%On target	88%On target	10%On target	19%On target	67%On target	$18
Oct 2026	82%On target	56%On target	85%On target	10%On target	16%On target	74%On target	$27

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	1,877	9.1M	93.6M	6.6M	2.3M	$134
Opus 5	2,375	17.6M	113.4M	8.4M	3M	$273
Opus 4.8	2,178	23.4M	93.6M	7.5M	2.9M	$283
Opus 4.6	4,612	71.9M	164.4M	15.1M	6.4M	$696
Opus 4.5	241	4.2M	8.1M	787K	351K	$39
Sonnet 5.5	5,183	29.1M	288.3M	20.3M	7.7M	$214
Sonnet 5	8,821	59.8M	382.2M	28.2M	10.9M	$564
Sonnet 4.6	8,347	126.6M	337M	29.6M	12.3M	$776
Sonnet 4.5	650	14M	26.8M	2.6M	1.2M	$78
Haiku 5.5	6,824	39.3M	377.6M	26.6M	12.6M	$17
Haiku 4.5	6,364	70.2M	282.3M	22.5M	9.2M	$172
GPT Realtime (meetings)	80	986K	3.2M	269K	115K	$8
Total	47,552	466.2M	2170.7M	168.3M	68.9M	$3,254

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.