The Training Log vs the Power Meter

The short version: Straude

Straude is a privacy-first activity tracker for AI-assisted coding — the Strava model applied to token telemetry. Its bundled ccusage release turns local agent logs (Claude Code, Codex, Gemini CLI, Qwen, Grok Build CLI) into a training log: daily spend, token volume, models used, session counts, streaks, and shareable public profiles. Prompts, conversations, and code never leave your machine. It is a clean take on the consistency question — and the practice-history framing is genuinely good.

SigRank reads the same four token pillars and asks the harder question: did the cascade compound? Υ Yield = cache_read × output / input² separates the session that built on cached context from the one that re-sent it. Straude keeps the streak alive; SigRank tells you whether the streak was worth keeping.

Feature comparison

Consistency is not efficiency

Straude's insight is that AI-assisted coding is a practice — and practices deserve logs, streaks, and pace. Correct. What a practice log cannot capture is session quality: two operators can both maintain a 30-day streak while one compounds cached context into high-yield output and the other burns fresh input every turn. The streaks look identical. The cascades are opposites.

SigRank's metric stack reads the shape Straude's log records: Υ Yield for the compounding question, Leverage for how much cache amplifies input, Velocity for throughput, SNR for signal density — plus the archetype that describes your operating pattern and the class tier that calibrates it to scale. In fitness terms: Straude is the activity feed, SigRank is the power meter and the VO2 max.

From practice to evaluation

If Straude already tracks your sessions, adding the evaluation layer is one install:

npm install -g sigrank
sigrank enroll      # create your operator identity
sigrank submit      # reads logs, scores, signs, publishes

Or paste your token counts into the /score calculator for an instant Υ Yield, class tier, and compression ratio.

Frequently asked questions

Is Straude a SigRank alternative?

They measure different axes of the same practice. Straude is the training log: it tracks whether you coded with AI today, your pace, your streaks, your history — the Strava model applied to token telemetry. It is privacy-first, ccusage-based, and genuinely well-positioned for the consistency question. SigRank is the evaluation layer: for each session, how efficiently did the token cascade compound? Straude answers 'did you train?'; SigRank answers 'how well did you drive?' The best answer is both.

Why compare a practice tracker to an evaluation system?

Because they sit on the same telemetry and solve adjacent problems. Straude's streaks and pace metrics tell you about your practice consistency — genuinely useful, and the reason the Strava model works for fitness. But a streak measures showing up, not driving well. SigRank's Υ Yield (cache_read × output / input²) measures the quality of each session's cascade: whether cached context compounded into output or fresh input burned without leverage. Consistency gets you on the road; efficiency tells you how fast you are going.

Can I use Straude and SigRank together?

Yes — and they complement each other well. Straude keeps your practice honest: streaks, daily pace, and the history that proves you are doing the work. SigRank scores the work itself: Υ Yield, Leverage, SNR, Velocity, your archetype, your class tier, and your rank against the field. The training log plus the power meter.

Which is better for proving operator skill?

SigRank — streaks prove dedication, not efficiency. An operator can hold a 30-day streak while re-sending context wastefully every session; the streak looks identical either way. Υ Yield distinguishes the compounding cascade from the burning one. Straude is the better practice companion; SigRank is the better evidence of skill.