Proxon

My Usage over time

Daily AI activity across approved tools, prompts, reviews, and workflows over the last year.

CURRENT STREAK

30days

ACTIVE DAYS

284

AI WORK SESSIONS

1,609

1,609 AI work sessions in the last year

BEST STREAK 30 DAYS

JunJulAugSepOctNovDecJanFebMarAprMayJun

MonWedFri

Captured from approved AI tools, prompts, reviews, and workflow runs.

LessMore

TOKEN TRACKING33.6M tracked tokens

Input, output, and cache reads across Claude, ChatGPT, Cursor, and Gemini sessions.

Input12.6M

Prompts, files, repo context

Output3.01M

Model responses and generated diffs

Cache18.0M

Reusable context served from cache

Cache share54%

COST ESTIMATES$97.17 estimated

Model-aware spend calculated from each tool’s measured tokens and approved internal rates.

ClaudeClaude Opus 4.5 · 326 sessions

$49.39

Deep coding and review · 12.6M tokens

ChatGPTGPT-5.4 · 602 sessions

$23.80

General drafting and analysis · 10.9M tokens

CursorClaude Sonnet 4.5 · 471 sessions

$20.67

Fast edits and synthesis · 7.44M tokens

GeminiGemini 2.5 Pro · 210 sessions

$3.31

Long-context research · 2.71M tokens

Cache savings estimate$62.08

AI Skill Level

A practical progression from first safe use to reusable AI workflows that learn from governed corporate memory. You are currently operating at L2 Contextual Work.

L0New

Safe first use

Knows what work is safe for AI, what data should stay out, and which starter tasks are useful.

L1Chat

Ad-hoc assistance

Uses AI as a conversational helper for summarizing, drafting, rewriting, and explaining work.

YOU ARE HERE

L2Contextual Work

AI works inside artifacts

Brings files, repos, docs, sheets, or project context into the session so AI can make bounded changes.

L3Orchestrate

Parallel and adversarial AI

Splits work across researcher, drafter, critic, and reviewer roles before final human approval.

L4Automate

Reusable workflows and triggers

Turns repeated AI work into governed workflows with triggers, permissions, approval gates, and outcomes.

L5Loop

Recursive improvement system

Closes the agentic loop by feeding outcomes, exceptions, and human feedback into corporate memory so every run improves the next.

Fluency Benchmark

Your AI fluency compared to similar Initech employees and industry peers in product and engineering IC roles.

PEERS38 peers

COHORTProduct + Engineering ICs

MATCHInitech employees

Context-Rich SessionsShare of AI sessions where you attach files, docs, repos, or examples before asking for output

p10p25p50p75p90

YOU68%

MEDIAN52%

76.8PCTL

Review Loops / WeekModel critique, second-pass review, or adversarial checks used before sharing final work

p10p25p50p75p90

YOU8.6

MEDIAN5.9

78.2PCTL

Reuse ConversionRepeated tasks converted into saved prompts, shared patterns, or workflow drafts

p10p25p50p75p90

YOU22%

MEDIAN18%

59.7PCTL

PROXON Coach Index · Q1 2026 (Initech internal peer set)

Next Things to Try

Concrete next moves and available resources that should move you from L2 into reliable L3 orchestration, then toward L4 automation and L5 recursive improvement.

NEXT MOVES3 actions

L3Run a review council on the next stakeholder artifactUse a drafter, critic, and editor pass before sending work outside the team.Fewer missed risks

L3Make source gaps explicit before every final draftAsk the model to list unknowns, missing evidence, and human confirmations before it rewrites.Cleaner approvals

L4Convert the QBR follow-up pattern into a workflowSave the prompt, define required inputs, add the CRM trigger, and keep a human approval gate.+7 hrs/mo

RESOURCESbest fit

Codex repo change patternAI Catalog · Engineering

L294%

Bot review councilAI Catalog · Product

L389%

Sales follow-up workflowAI Catalog · GTM

L486%

Source-gap checklistCoach template

L384%

My Achievements

Badges for the different types of AI work you have already done, plus the next badges you are close to earning.

Mar

Context Loader

Attached real source material in 18 sessions this month

Feb

Diff Reviewer

Asked AI to make bounded artifact changes and explain the diff

Jan

Policy-Safe Operator

Kept customer data out of unapproved prompts across every session

Jan

Prompt Sharer

Shared a reusable account-summary prompt with the product team

Trial

Review Council

Ran critic and editor passes before a customer-facing QBR follow-up

Workflow Candidate

Identified a repeat QBR follow-up pattern ready for automation