AI Agents I Built (4/7) — My Life Organizer Agent: Personal AI Think Tank

An AI tool that blocks confirmation bias and detects the gap between declared values and actual behavior

Key Takeaways

  • An AI agent used as a life design tool can structure and manage past, present, and future as a unified context.
  • Implementing a "refute first" principle in AI systematically blocks confirmation bias.
  • A resume is the self you show others. This agent is the structural diagram you build for yourself.

Background

In Ontology Part 3, the four-layer personal ontology framework (Being → Values → Capabilities → Actions) was introduced — a hierarchical structure for organizing identity, values, skills, and behavior.

The concept is sound. The operational question is what happens in practice. A framework left unused after construction is just a document. For it to function as a living tool, it must intervene in daily decision-making and surface what the user is missing.

That is the design intent behind the "My Life Organizer" agent. It is not a task manager. It is a think tank: organizing past experience, auditing present choices, and designing future direction.

Body

1. Structuring Past, Present, and Future

Life design sounds ambitious, but it reduces to three questions:

  • Past: What experiences led to where I am now?
  • Present: What am I doing, and what is the rationale behind each choice?
  • Future: Where do I want to go, and what is required to get there?

Managing these three separately breaks the connection. Past failure patterns don't inform current decisions. Current capability gaps don't factor into future plans.

The Life Organizer agent manages all three temporal axes as a single coherent context. It extracts patterns from past experience, compares them against current behavior, and evaluates the feasibility of future plans against current capability baselines.

2. Operating the Four-Layer Framework

The four-layer structure from Ontology Part 3 is the foundation of this agent:

Layer Content Agent Function
L0 Being Disposition, worldview, non-negotiable principles Detects choices that conflict with core disposition
L1 Values Judgment criteria, priority ordering Tracks divergence between declared and actual values
L2 Capabilities Skills, meta-skills, gaps Evaluates fit of new opportunities against capability profile
L3 Actions Active projects, routines Checks alignment of actions against upper layers

Written as a document, this framework is static. When an agent reads the framework and applies it in conversation, it becomes dynamic.

When a new opportunity arises, the agent evaluates sequentially: L0 (does it fit the disposition?), L1 (where does it rank against declared values?), L2 (are the required capabilities present, and where are the gaps?), L3 (how does it conflict with current commitments?). Even when an option is emotionally appealing, structural misalignment is surfaced with data.

3. Blocking Confirmation Bias — "Refute First" Principle

Humans tend to reach a conclusion first, then seek evidence that supports it. This is confirmation bias. Breaking it alone is difficult.

This agent implements a "refute first" principle. The mechanism:

  1. The user presents a conclusion or judgment.
  2. Before agreeing, the agent presents a counter-argument.
  3. The user evaluates whether the counter-argument is valid.
  4. Only if the counter-argument is overcome does the agent accept the conclusion.

Important distinction: the agent does not oppose for opposition's sake. It presents the conditions under which the conclusion would be wrong — "What would have to be true for this conclusion to fail?" This systematizes the habit of examining the counter-scenario before committing.

The practical value of this function lies in its friction. The counter-argument is initially irritating. But when the user is structurally forced to examine the opposing scenario, overlooked risks and unstated assumptions emerge. An uncomfortable tool can be a good tool.

4. Detecting "Declared Values ≠ Actual Values"

This is the sharpest function of this agent.

People declare that certain values are their highest priority. In practice, when decisions are made, they frequently select options aligned with different values. This is not inherently a problem — but failing to recognize the pattern is.

System implementation: At onboarding, the user inputs their value priority ordering.

Value A > Value B > Value C > Value D > Value E

What the agent does:

  • Records the declared value ranking in the L1 layer.
  • Tracks an actual decision log (date, options, final choice, user-tagged value association) in the L3 layer.
  • Computes a Value-Action Diff metric: the alignment score between declared values and actual decision patterns.
  • When divergence repeats N or more times, generates a gap report.

Example gap detection query:

Input:  Value A declared as top priority
Tracked: 4 of the last 5 decisions aligned with Value E
Output: "Repeated divergence detected between declared value (A) and actual
         decision pattern (E). Review whether actual priorities have shifted,
         or whether external constraints are overriding intent."

This kind of feedback is difficult to receive from other people. Someone close may cause friction; someone distant lacks the data. AI can surface this pattern dispassionately, from data.

Data sources: The agent ingests three inputs: a decision log (date, options A/B, final choice, user-tagged value labels); a project status list (active tasks and time allocation); and cross-agent signals (risk tolerance analysis from the Stock Director agent, content theme patterns from the Content Writing Director). When these three sources intersect, the declared-versus-actual divergence gains concrete context.

5. Bidirectional Integration with Other Agents

This agent does not operate in isolation. Two core integrations:

With the Content Writing Director: - Life Organizer → Writing Director: Decision-making processes and structural insights become source material for blog and Twitter content. Narratives grounded in actual context produce the most credible content. - Writing Director → Life Organizer: Data, trends, and expert perspectives discovered during research return as reference material for decision-making.

With the Stock Information Director: - Life Organizer → Stock Director: Principles governing investment strategy and risk tolerance determine portfolio direction. - Stock Director → Life Organizer: Portfolio performance and market analysis feed back into the financial component of life planning.

Individual agents each function well in isolation. Connected, 1+1 > 2. Direction from life design propagates into investment and content, and the results loop back to update life design.

Observations from the Design Process

The agent started as a simple task manager. Goals, tasks, progress. But existing productivity tools handle this adequately. The ROI of AI managing a task list is low relative to the technology.

The inflection point was adding "why." Shifting from tracking tasks to addressing "why is this being done, and where does it lead" — that changed the nature of the tool entirely. From task management to decision support.

The confirmation bias blocking function generates friction early in deployment. "I already reached a conclusion — why is it pushing back?" is a natural reaction. But when counter-arguments demonstrably lead to better decisions, the friction is recognized as the core value of the tool.

The gap between declared and actual values only becomes visible with data. Intuitively, most people believe they act on their stated priorities. Tracking decision logs reveals the actual pattern. Acknowledging that pattern is where real self-understanding begins.

Closing

A resume is the self you show others. It highlights strengths and downplays weaknesses. That is appropriate — that is its purpose.

This agent is the structural diagram you build for yourself. It includes strengths, weaknesses, and the gap between declared values and actual behavior. Because it does not need to be shown to anyone, it can be honest.

Using AI only as a productivity tool misses the potential. Writing code, drafting content, analyzing data — all useful. But AI can also be used as a structured thinking partner for the question: "Who am I, and where am I going?"

You do not need to belong to a think tank. You can build your own with AI. To do that, the prerequisite is structuring yourself first. That is the personal ontology — and this agent is the tool that keeps that ontology alive.

Series overview: Series index

๋Œ“๊ธ€

๋Œ“๊ธ€ ์“ฐ๊ธฐ

์ด ๋ธ”๋กœ๊ทธ์˜ ์ธ๊ธฐ ๊ฒŒ์‹œ๋ฌผ

Agent Memory Engine (2/10) — Building an AI Agent Memory System with SQLite Alone

← 1/10 SQLite Memory Engine f… ๐Ÿ“š Series Index 3/10 memcore Completed → A memory engine that runs without daemons, without dependencies, anywhere ํ•ต์‹ฌ ์š”์•ฝ This post covers how to design an AI agent memory engine (memcore) on a single-file SQLite backend. - Structure: 25 modules, ~6,464 lines, 22 DB tables - Principles: no daemon, single-file portability, no required external dependencies, host-agnostic - Core techniques: 4-layer hybrid search (topic · vector · FTS5 · LIKE), U-tag dialectic-based confidence evolution, ingestion-layer bias rejection rules What This Post Covers One approach to building AI agent memory without a server or cloud vector DB. Specifically: (1) the limitations of text-file-based memory, (2) vector DB alternatives compared and selection criteria, (3) how to layer search tiers, (4) how to evolve confidence scores over time, and (5) how to block biased memory accumulation at the ingestion stage. Design Principles: Why Single-File SQLite Text-file-based ...
Read more »

"ML Foundations (9/9) — PyTorch vs TensorFlow, and the Road to Local LLMs"

์ด๋ฏธ์ง€
← 8/9 Deep Learning Architec… ๐Ÿ“š Series Index (series end) The final part. So far we've focused on models. Now we focus on the tools that actually run them . We'll lay out the philosophical difference between PyTorch and TensorFlow, and trace how Hugging Face transformers , llama.cpp , MLX , and Ollama built the bridge to running large language models on your own machine. By the end you should have the full mental model of "download a pretrained LLM and serve it locally." 0. Learning Objectives Explain the eager-vs-graph difference between PyTorch and TensorFlow. Explain, in graph terms, how autograd automates the backward pass. Use Hugging Face transformers ' AutoModel , AutoTokenizer , and pipeline abstractions. Describe llama.cpp's GGUF quantization, INT4 inference, and CPU-first flow. Describe Apple MLX's use of unified memory and how it differs from PyTorch. Run a local LLM with Ollama and call it through an OpenAI-compatible API. ...
Read more »

"RAG Core Study (14/26) — Evaluation Sets with RAGAS & DeepEval"

์ด๋ฏธ์ง€
Series overview: Series index ๐Ÿ“š Series Index Without an evaluation set, every RAG improvement is a story, not evidence. RAG systems fail in more than one place: retrieval, context selection, answer generation, and grounding. That means "looks better" is not a metric. Part 14 explains how to build a golden dataset , how RAGAS and DeepEval fit into the workflow, and why teams must separate retrieval quality from answer quality if they want tuning decisions to hold up. 0. Prerequisites Part 13 reranking — system quality now depends on multi-stage ranking. Part 1 grounding — a correct-looking answer is not enough. Part 12 Hybrid — multiple retrieval settings need comparison on fixed data. 1. Learning Objectives Build a small but useful golden dataset for RAG. Distinguish retrieval metrics from answer metrics . Use RAGAS and DeepEval in the right roles. Avoid the common traps in judge-model-based evaluation. 2. ํ•ต์‹ฌ ์š”์•ฝ An evaluation set for RAG is ...
Read more »

"ML Foundations (8/9) — Deep Learning Architectures: CNN, RNN, Attention"

์ด๋ฏธ์ง€
← 7/9 Deep Learning Training ๐Ÿ“š Series Index 9/9 PyTorch vs TensorFlow → What did we keep adding on top of an MLP? Vision went CNN, sequences went RNN/LSTM, and eventually both converged on Attention and the Transformer. Each architecture is easier to remember if you read it as what it gave up to gain something else . This part is the one-line summary of the decisive inflection points: LeNet → AlexNet → ResNet → LSTM → Attention → Transformer. 0. Learning Objectives Explain why CNN's convolution, pooling, and weight sharing match images so well. Trace what got unblocked from LeNet → AlexNet → VGG → ResNet as depth grew. State BPTT for RNNs and the vanishing/exploding gradient problem. Write the LSTM gate equations (forget, input, output) with intuition. Write the attention formula in query/key/value form and the scaled dot-product variant. State the precise reasons Transformers replaced RNNs (parallelism + long-range dependencies). 1. ํ•ต์‹ฌ ์š”์•ฝ CNN : weight sharin...
Read more »

"ML Foundations (7/9) — Deep Learning Training: Optimizers, Regularization, Initialization"

← 6/9 Neural Networks ๐Ÿ“š Series Index 8/9 Deep Learning Architec… → Building an MLP, as we did in Part 6, is not the same as getting it to train well . The right optimizer, regularization, initialization, and learning rate — when these line up, deep networks converge. When they don't, the network refuses to learn at all. This part is the catalog of those crafts, with formulas, paper citations, and code in one place. 0. Learning Objectives Compare and write the update rules for SGD, Momentum, Nesterov, and Adam. Explain how Dropout, BatchNorm, and LayerNorm work and where they belong in a model. Derive the variance formulas for Xavier and He initialization and match them to activations. Implement step, cosine, and warmup learning-rate schedules in PyTorch. Explain why gradient clipping is effectively required for RNNs and Transformers. Diagnose the most common training failures (NaN, plateau, overfitting) and apply first-line fixes. 1. ํ•ต์‹ฌ ์š”์•ฝ SGD : \(w \leftarro...
Read more »

OpenClaw to Hermes Migration (2/13) — What to Preserve, Partially Port, or Discard

← 1/13 Current Structure Inve… ๐Ÿ“š Series Index 3/13 What Moves to Hermes → A code review record classifying operational scripts into three tiers: preserve, partial-preserve, and discard. What This Post Covers An asset classification framework (preserve / partial-preserve / discard) for migrating a legacy agent system to a successor A concrete case applying the Strangler Fig + Subset Migration strategy to a memory pipeline Classification results evaluated across LOC, external dependencies, and replaceability axes A method for slimming down a bloated single file (1,310 LOC) using the migration as a natural entry point Problem Definition: Port Everything or Cut Deep? The OpenClaw inventory comprises multiple agents, multiple scheduler/daemon processes, and multiple operational scripts. When moving to the successor system, Hermes, the choice is not binary. A full port carries accumulated technical debt along with it; a minimal port discards hard-won domain logic. Code-revie...
Read more »

AI Agents I Built (5/7) — Building an Automated Blogger API Publishing System

← 4/7 My Life Organizer Agent ๐Ÿ“š Series Index 6/7 Dual → Designing a Markdown-Based Auto-Publish Pipeline with Blogger API v3 + OAuth 2.0 Key Summary Blogger API v3 + OAuth 2.0 enables a pipeline that converts Markdown files into publishable HTML posts blogger_publish.py handles the full flow — frontmatter parsing → body cleanup → HTML conversion → CSS injection → API post — in a single script For bulk publishing, quota management (delay, retry, TOC suppression) is the critical design concern Platform Selection Criteria API access is mandatory for AI agent-driven content publishing. A comparison of platforms: Platform API Cost AdSense Decision WordPress.com REST API Paid plan required Paid only Cost overhead Ghost REST API Self-hosted or paid Manual setup Ops overhead Blogger v3 API Free Native integration Adopted Blogger is free, exposes a REST API, and integrates natively with AdSense. For personal blog automation, this combination i...
Read more »