How do I measure personal growth without comparing myself to others?

Track Your Own Growth With AI Memory, Not Peers

Use Claude's persistent memory (rolled out March 2026, updated July 2026) or ChatGPT's GPT-5.6 Dreaming Architecture to log weekly facts about yourself, then ask it monthly to compare you to your prior self, not peers, the exact instruction Brad Stulberg and Steve Magness give in The Passion Paradox: judge yourself against earlier versions of yourself, not others.

"How do I measure personal growth without external benchmarks" is one of the more honest questions people type into a search bar, because the real problem underneath it usually isn't measurement, it's that every visible benchmark is someone else's. LinkedIn is a highlight reel, a founder peer group is a comparison engine disguised as a support group, and most habit-tracking apps quietly gamify you against a leaderboard whether you asked for one or not. I get some version of this question from coaching clients almost monthly, usually right after they've had a bad week and gone looking for proof they're behind.

The tool I actually recommend now isn't a new app, it's a feature most people already have and mostly use wrong: the persistent memory built into Claude and ChatGPT. Anthropic rolled out cross-conversation memory to every Claude user, including the free tier, in March 2026, and updated it in July 2026 so it stores individual categorized entries rather than one running daily summary, which matters because it means you can ask it to retrieve a specific thread of your own thinking from months ago instead of a vague paraphrase. You manage it under Settings, Capabilities, Memory, and there's an Incognito mode for anything you don't want remembered. OpenAI's equivalent, the "Dreaming Architecture" memory system that shipped with GPT-5.6 in June 2026 (GPT-5.6 itself went public on July 9, 2026), does something similar, synthesizing patterns across a multi-year horizon rather than a single session, with a January 2026 update that improved reliable recall of details from over twelve months back.

What makes this useful for the specific question of growth-without-benchmarks isn't the storage, it's what Brad Stulberg and Steve Magness argue in The Passion Paradox, a book about the biology of drive rather than productivity software. Their instruction is blunt: "Don't judge yourself against others. Judge yourself against prior versions of yourself and the effort you are exerting in the present moment." They tie this to what they call the mastery mindset, one principle of which is being the best at getting better, not the best, period, a distinction that only works if you actually have a record of your prior self to compare against. Most people don't. Their memory of six months ago is a vague, mood-colored reconstruction, not the actual reasoning they used at the time. An AI with persistent memory is, for this narrow purpose, a more honest record than your own recollection, because it isn't editing the story for you the way Stulberg and Magness describe our brains doing after setbacks.

Here's the protocol I actually run, which I call the prior-self protocol, four steps, weekly. First, once a week, tell Claude or ChatGPT three specific facts to remember: one decision you made, one honest read of how you felt making it, and one number, a metric, a word count, a number of client calls, whatever's real for that week. Second, monthly, ask it explicitly to compare this month's entries to what it has stored from one, three, and six months back, and force the comparison to be about your own trajectory, not phrased as "how am I doing," which invites generic encouragement instead of a real pattern. Third, ban comparison language from your own prompts entirely, no "how do other founders handle this," because that's the exact door that lets external benchmarking back in through the AI itself. Fourth, quarterly, review what's stored and delete anything stale, because memory that isn't pruned drifts into a caricature of you rather than an accurate one.

The honest limit is that this isn't a substitute for real business metrics, and I tried using it that way for a few weeks before abandoning it, memory-based retrieval is good at surfacing how you thought about a decision, not at replacing a revenue spreadsheet or a proper analytics dashboard. It's also opt-in and imperfect: both companies' memory systems occasionally misremember or drop details, and you should assume anything genuinely sensitive doesn't belong in it regardless of the incognito option. Used narrowly, for the specific job of proving to yourself that you're not the same person you were three months ago, it's the best tool I've found, mostly because it isn't trying to be a benchmark at all.

Sources: Anthropic, Claude memory release notes and support documentation (2026); OpenAI, "Memory and new controls for ChatGPT" and GPT-5.6 documentation (2026); Brad Stulberg & Steve Magness, The Passion Paradox.