Support

How Learning Works

Tempreon learns how you think and work through two systems — Tempering (passive) and Honing (active).

The Learning System

This is the core of what makes Tempreon different. While other tools forget you after every session, Tempreon actively learns your patterns and gets smarter over time.

The learning system has two modes:

Tempering (Passive Learning)

Tempering runs in the background. You don't need to do anything — just use your AI tools normally, and Tempreon picks up on patterns.

It works across three levels of depth:

Flash Tempering

Quick, same-session pattern detection. When you correct your AI's output or show a clear preference in how you want something done, Flash Tempering captures it immediately.

Example: You ask for a summary and tell the AI "That's too long — keep summaries under 3 sentences." Flash Tempering captures this as a signal.

Sustained Tempering

Multi-session pattern enrichment. When the same pattern appears across multiple sessions, Sustained Tempering validates and strengthens it.

Example: Across five sessions, you consistently restructure recommendations to lead with data before analysis. Sustained Tempering identifies this as a reliable process pattern.

Core Tempering

Cross-domain intelligence synthesis. When a pattern holds true across different Domains (projects, contexts, roles), Core Tempering elevates it to a core part of your profile.

Example: Whether you're working on marketing or engineering, you always prefer bullet-point summaries over paragraphs. Core Tempering recognizes this as a fundamental communication preference.

Proving

The validation layer. Patterns are checked against your Core Imprint for consistency. This prevents the learning system from capturing one-off behaviors as permanent preferences.

Honing (Active Learning)

Honing is where you actively participate in the learning process. It has three expressions:

Sharpening Questions

Tempreon asks you specific questions to resolve ambiguity in what it's learned. These appear in the Sharpening section of your dashboard.

Example: "You've used both formal and casual language in client communications. Is the difference based on the client, the stage of the relationship, or something else?"

Your answers sharpen the system's understanding significantly — a single clear answer can be worth dozens of passive observations.

Corrections

When your AI gets something wrong, say so. Corrections are the highest-value learning signal:

  • "That's not how I'd phrase it — I never use the word 'synergies'"
  • "Don't suggest we rush this. I always prefer getting it right over getting it fast."
  • "Actually, I want the opposite — lead with the recommendation, not the analysis."

Each correction adjusts the relevant Instinct immediately.

Focus Areas

The system identifies areas where it needs more data and guides learning toward those gaps. These appear on the Sharpening page as prioritized topics.

What Gets Learned: Instincts

The output of all this learning is Instincts — validated behavioral patterns organized into three types:

TypeWhat It CapturesExamples
StyleCommunication, tone, formatting"Prefers short paragraphs", "Never uses emoji in business writing"
ProcessWorkflow, tools, approach"Reviews data before making recommendations", "Prefers to draft → edit → finalize"
JudgmentDecisions, trade-offs, priorities"Prioritizes speed over perfection for internal docs", "Always considers cost of being wrong"

More detail in the Instincts docs.

Tier Differences in Learning

CapabilityFreeStandardPremium
Flash TemperingLow frequencyDailyContinuous
Sustained Tempering--DailyContinuous
Core Tempering----Continuous
Proving----Continuous
CorrectionsYesYesYes
Sharpening Questions--YesYes
Focus Areas----Yes
Instinct visibilityLimitedFullFull + Forge Score

Corrections are available on all tiers because trust is fundamental — you should always be able to tell your AI when it's wrong.

Try Saying...

You can interact with the learning system naturally:

  • "Remember that I prefer data-driven recommendations over opinion-based ones" — Direct preference storage
  • "That's not my style — I never use exclamation points in professional writing" — Correction that feeds learning
  • "How well do you know my communication preferences?" — Check learning depth
  • "What patterns have you noticed about how I work?" — Surface learned Instincts

Use Case: The AI That Gets Better Every Week

Nina is a content strategist who uses Claude Desktop daily. In week 1, she had to constantly remind Claude about her writing style — no jargon, short sentences, always include a clear CTA. By week 3, Tempreon's Tempering system had captured these as validated Style Instincts. Claude started applying them automatically. By month 2, even her Process Instincts were dialed in — Claude knew to outline before drafting, to propose 3 headline options, and to include SEO considerations without being asked. Nina's editing time dropped by 60%.