Disclaimer: I do not promote any of the tooling mentioned in this blog post nor have I used the vast majority of them. All of the text written by me is my personal opinion. All of the text generated by LLM is its creation – I will be explicit where it is LLM and where it is me.

Unlike most posts where I make emphasis on my own experience and writing by myself, this blog post will be assisted with LLM, but in a good way. I want to create a comprehensive AI toolset list to explore for myself. Although it will have some tools very specific to my personal needs (like climbing) I believe you can apply the same approach to get your ideal list. This could have been LinkedIn post “Definite list of AI tools you must use” but I cannot stand those posts. This is just idea on how you can come up with your own list:

Step 1: Refine research plan

Prompt: “Refine this prompt: Perform research on the landscape of currently available LLM products and AI tooling for both professional and personal development for the user profile and wider audience, using referenced examples of already used tooling. User Profile: A Software Engineer at a big tech company in the Seattle area. Family with two kids, sports (climbing, Muay Thai, running), writing blog posts, coding for fun, and wondering about life in general. Researching topics such as: health, finance, 3d printing, writing about tech, coding, taxes, parenting, relationships, career decisions, trip planning, real estate, etc. Reference examples:  ChatGPT, Gemini (deep research, chat, image gen), Claude Sonet for coding, locally running small LLMs, trying out different AI startups for intermittent tasks.”

The above text was me. The below text is the proposed research plan by LLM:

(1) Research the current landscape of major LLM products (such as ChatGPT, Gemini, Claude) and emerging AI startups in late 2025, specifically analyzing their capabilities for professional software engineers (coding agents, debugging, architecture design).
(2) Investigate the ecosystem of open-source and locally executable LLMs (compatible with Ollama) and privacy-focused tools suitable for a tech-savvy user running consumer hardware.
(3) Search for specialized AI tooling and applications tailored to the user's specific hobbies and creative interests:
(a) Sports performance analysis for climbing
(b) Sports performance analysis for Muay Thai
(c) Sports performance analysis for running
(d) 3D printing modeling and slicing optimization
(e) Technical blogging and creative writing assistance
(4) Find AI resources and tools assisting life management, covering:
(a) Parenting, education, and kids' activities in the Seattle/Bellevue area
(b) Real estate market trends and housing analysis in Vancouver, BC
(c) Personal finance, tax preparation, and investment tracking in the Seattle/Bellevue area
(5) Explore AI interfaces designed for broad life topics, such as philosophical conversation, relationship coaching, career development planning, and travel logistics.
(6) Synthesize the findings to categorize tools by professional utility versus personal development, comparing them against the user's current stack (ChatGPT, Gemini, Claude, Ollama) to identify gaps and superior alternatives.

Step 2: Run multiple researches and synthesize takeaways

I’ve run 4 reports with the above prompt or some variations of it. In the end I got four 20 page documents. After skimming through them it is obvious that either: a) some info is outdated, like recommending Claude Sonet 3.5 when I’m already using 4.5; b) recommendations are too generic; c) some sections were too verbose and some too short for my needs; d) some recommendations are contradictory, for instance using Ollama for privacy and then suggesting some startups that suggest uploading private data to small startups.

One idea I had was to feed all documents into another research specifically asking for summary with bullet points. Unfortunately the result was a 5th document of not better quality. Another approach I took was: “Create a maximum 5 pages summary from the four similar documents. Do bullet points. Keep links to websites. Do NOT do research.” This gave much better results.

Step 3: Edit the result

So I’ve got approximately what I need. The next step was to iterate a few times on the list. E.g. I asked LLM to add a list of general health and longevity tooling following the same format. Copy-paste, read through, and add more sections. The appendix text in the end is the list by AI, with minor edits from me.

Step 4: Action Plan

It is obvious that I won’t be trying all of the tooling (that would be crazy) but to do exploration of what’s available and within my area of interest. As an action plan I highlighted some tooling to use and play with below or use more actively:

  • llms.txt – a file to be added to this blog post so AI knows how to read it.
  • Ollama: The CLI standard for running local models. Already using it, but probably use even more. (got M4 apple processor with 32Gb memory so some LLMs run just fine).
  • NotebookLM: Google’s AI-Powered research partner. Have seen demos of this one and should try.
  • Belay AI: analyzing center of mass and hip trajectory when climbing. Didn’t know such a thing existed. Would be curious to try out.
  • Garmy: Python library and MCP server linking Garmin data to Claude Desktop or Cursor for agentic analysis. Sounds like something I would like to play with next.
  • Meshy: Converting prompts to 3D models. Already tried but wan’t too happy with the results. Will give it another try.
  • Layla & Wanderlog: AI travel agents. Used general LLMs for my travel planning before but will be curious to try tailored tool.
  • Orai: Pocket AI coach analyzing recordings for filler words, energy, and clarity.

Conclusion

I believe the AI and LLM tooling landscape is very saturated. There is a tool or a startup for almost anything you can think of. The point is not the specific list but how I arrived at it and how it is tailored to my needs. In this blog post I provided a method at arriving at your own list of AI tooling that is applicable specifically to you.

Alert: wall of text below.

APPENDIX: THE RESULT

Disclaimer: I am not advertising or promoting any of the tooling below, have no affiliation to any of the companies or products mentioned. The text below is generated by LLM. I only reviewed it.

Executive Summary: The Shift to Agentic Workflows

The technological paradigm has shifted from “Chatbots” (passive Q&A) to “Agents” (active execution). The competitive advantage in 2025 belongs to the “Augmented Architect” who orchestrates specialized AI entities to manage full-stack development, complex financial engineering, and physical performance.

1. The Professional Engineering Workbench

The modern workflow bifurcates into Integrated Agents (living in the IDE) and Headless/Terminal Agents (operating autonomously).

The Battle for the IDE: Cursor vs. Windsurf vs. Copilot

  • Cursor: The “Architect’s Instrument”.
    • Best For: Heavy refactoring and legacy codebases.
    • Key Feature: “Composer” and “Shadow Workspace” index the entire codebase to predict multi-line edits and handle global refactors.
    • Model: Uses Claude 3.5 Sonnet for superior code structure nuance.
    • Pricing: $20/month; generally preferred by power users over Copilot.
  • Windsurf: The “Flow State” Facilitator.
    • Best For: Greenfield projects and rapid prototyping.
    • Key Feature: “Cascade” flow tracks user actions (terminal commands, clipboard) to infer intent, actively running tests and fixing errors.
    • Differentiation: Focuses on keeping the developer in “flow” rather than granular control.
  • GitHub Copilot: The Enterprise Incumbent.
    • Best For: Corporate environments with strict compliance needs.
    • Status: Now includes “Agent Mode,” but critics note the chat often feels disconnected from the editor compared to AI-native rivals.

Command Line & Autonomous Agents

  • Claude Code (CLI): A terminal-resident agent that replaces the chat interface. It can navigate directories, read files, execute Unix commands, and handle large-scale refactoring.
  • Cline / Roo Code: Open-source VS Code extensions that act as “Headless Developers.” They can execute terminal commands and create files autonomously, allowing for a “Bring Your Own Key” (BYOK) model.
  • Deep Think Models: Google’s Gemini 2.5 Pro utilizes “parallel hypothesis testing,” allocating a “thinking budget” to simulate System 2 thinking for architectural reviews and debugging race conditions.

The Sovereign Stack: Local Inference on Apple Silicon

For IP protection and privacy, running models locally on M3/M4 Max chips (Unified Memory) is the standard.

  • Ollama: The CLI standard for running local models (like Docker for LLMs).
  • LM Studio: A GUI alternative for discovering and testing models.
  • Top Open-Source Models:
    • DeepSeek-Coder-V2: Uses Mixture-of-Experts (MoE) for high reasoning with efficient inference; ideal for logic-heavy tasks.
    • Qwen 2.5 Coder: The premier open-source choice for daily coding, rivaling GPT-4 in benchmarks (88.4% HumanEval) and running on 32GB+ RAM.

Career Engineering & Strategic Presence

Interview Intelligence (The “Copilot” Era)

  • Final Round AI: Real-time “Interview Copilot” offering transcription and live hints.
  • InterviewBee AI: Adaptive mock interviews that dynamically adjust difficulty.

Strategic Networking & Personal Branding

  • Supergrow: LinkedIn growth tool generating tone-matched content from past posts.
  • Taplio: Identifies viral technical topics and drafts high-visibility posts.

Performance Engineering (Automated Brag Docs)

  • Lattice AI: Auto-drafts reviews, translating engineering metrics into business impact narratives.
  • Fellow: Auto-generates “Brag Docs” by tracking wins from 1:1s year-round.

Public Speaking & Leadership Intelligence

Voice Cloning & Auditory Feedback

  • ElevenLabs Voice Cloning: Clones your voice to objectively audit delivery and identify awkward phrasing before speaking.
  • Orai: Pocket AI coach analyzing recordings for filler words, energy, and clarity.

Simulation & Real-Time Coaching

  • Yoodli: Simulation platform providing private analytics on eye contact and pacing during practice speeches.
  • Poised: Real-time meeting assistant offering live, private feedback on speaking speed and confidence.
  • VirtualSpeech: VR-based training for practicing presentations in realistic 3D environments (e.g., boardrooms).

2. Wealth Management

For a Seattle engineer with RSUs and complex taxes, standard budgeting apps are insufficient.

The “Finance as Code” Approach (Privacy-First)

  • Beancount: A Python-based double-entry bookkeeping system that stores financial records in plain text. It treats finance like code (version control, CI/CD).
  • Fava: The web UI for Beancount.
    • Fava Investor Plugin: Calculates IRR and tracks asset allocation across disparate accounts.
    • Automation: Python scripts (e.g., wash-sale-tracker) can parse trade history to track wash sales and automate RSU vesting tracking.

The SaaS Optimization Route

  • Prospect: Specialized for tech employees with ISOs/RSUs. It models tax implications of exercising options and calculates the “AMT crossover point” to prevent surprise tax bills.
  • Compound Planning: Tracks net worth across illiquid assets and models scenarios for RSU vesting and tax cliffs.
  • StockOpter: Specifically addresses equity compensation guidance and AMT modeling.
  • HouseSigma: Essential for the Vancouver, BC market. Uses AI to provide “Sold” history and valuation estimates, unlocking data previously gated by realtors.
  • VanPlex: Analyzes zoning data to identify “under-utilized” lots suitable for multiplex development (Bill 44), aiding in investment arbitrage.

Financial Education & Investment Intelligence

Tools for deep research, interactive learning, and risk-free simulation (Paper Trading).

Deep Research & Earnings Intelligence

  • AlphaSense: Institutional “semantic search” for broker research and earnings calls; highlights off-script management answers.
  • Quartr: Mobile access to live earnings calls/transcripts; “search across audio” finds specific keyword mentions instantly.
  • FinChat.io: “ChatGPT for Finance” providing sourced answers from verified 10-Ks/10-Qs to minimize hallucinations.

Interactive Education & Family Literacy

  • Magnifi: AI investing tutor; explains portfolio balance and stock fit via conversational interface rather than just charts.
  • Zogo: Gamified literacy app for families; breaks complex topics into bite-sized modules with rewards.

Simulation & Practice (Paper Trading)

  • Thinkorswim: Institutional-grade paper trading with “PaperMoney” to test strategies risk-free.
  • Webull: UX-friendly platform for beginners to practice trading mechanics before deploying capital.

3. The Quantified Athlete: Physical Intelligence

AI is transitioning from logging data to providing active biomechanical coaching.

Rock Climbing

  • Belay AI: Uses computer vision (pose estimation) on a smartphone to analyze center of mass and hip trajectory, identifying micro-inefficiencies in movement.
  • KAYA Pro: Digitizes climbing sessions and calculates “Workload” to prevent overtraining. It filters “beta” videos by body morphology (e.g., finding beta for a specific height).
  • Lattice Training: Uses datasets to benchmark finger strength and build periodized training plans.
  • Crimpd: Utilizes analytics to prescribe hangboard workouts and manage training loads.

Muay Thai & Running

  • Sensei AI: A virtual coach for Muay Thai that analyzes shadow boxing via camera. It corrects hip rotation on kicks and guard retraction.
  • RunDot: The data scientist’s choice. Uses “Environment Normalization” to adjust pace targets based on heat/humidity, ensuring constant physiological stimulus.
  • Runna: Focuses on UX and community, gamifying the training process for better adherence.

General Health & Longevity Intelligence

The intersection of “Quantified Self” and AI for preventative health and programmable biology.

Developer-Friendly Health Data

  • Garmy: Python library and MCP server linking Garmin data to Claude Desktop or Cursor for agentic analysis.
  • HealthGPT: Open-source iOS app using on-device LLMs to query Apple Health data privately.

Longevity & Preventative Analytics

  • InsideTracker: “Programmable” biology platform; integrates wearable APIs and uses Terra AI to map blood biomarkers to peer-reviewed optimization protocols.
  • Function Health: Deep clinical baseline with a 100+ biomarker panel for early detection and chronic disease prevention (closed system).
  • Superpower: Accessible longevity diagnostics and biological age tracking at a subscription-friendly price point.

4. Creativity & Fun

Bridging the gap between software and physical artifacts.

3D Printing Stack

  • OrcaSlicer: The “Open Source Victory” for 2025. Offers granular control (jerk/acceleration settings) and “Scarf Joint Seams” for aesthetics.
  • Obico: AI failure detection (spaghetti detective). It monitors the print bed via camera and pauses prints to prevent fire/waste. Can be self-hosted on a Raspberry Pi.
  • Zoo (formerly KittyCAD): “Text-to-CAD” API. Generates editable, parametric CAD models (code-based) rather than simple meshes.
  • Meshy: Generates 3D assets from text prompts, best for rapid prototyping or game assets.

Knowledge & Blogging Pipeline

  • Repo-to-Blog: A workflow using Gitingest or GitHub Actions to convert codebases into token-optimized summaries. These are fed into LLMs to automatically generate technical “DevLogs” from commit history.
  • Obsidian + Smart Connections: A “Second Brain” setup where the plugin uses local embeddings to allow you to “chat” with your notes vault.
  • llms.txt: A new standard for 2025. Adding this file to a personal site makes it indexable by AI agents.

Music Exploration & Vibe Matching

  • Spotify AI Playlist: Generates playlists from creative text prompts (e.g., “songs for a rainy cafe”).
  • PlaylistAI: Creates playlists from text prompts, images, videos, or festival posters.
  • Maroofy: Search engine that matches songs by “audio vibe” rather than just artist similarity.
  • Cyanite: Advanced “Free Text Search” to find songs matching specific moods or descriptors.
  • Music-Map: Visual tool that creates a floating “cloud” of related artists based on fan affinity.
  • NotebookLM: Google’s research tool; upload your concert history/venue calendars to create a custom event finder.
  • HyperWrite: AI agent capable of browsing the web to find specific tickets or venue schedules for you.

5. Life Logistics

Tools acting as “Chief of Staff” for the household.

  • Ohai.ai: Ingests unstructured data (screenshots of flyers, voice memos) to manage family calendars and conflicts.
  • Milo: An SMS-first family assistant (powered by GPT-4) that manages logistics via natural conversation.
  • HomeZada: A digital home management platform that uses AI to predict maintenance costs, schedule seasonal repairs.
  • Familymind: Synthesizes school PDFs and sports schedules into a master calendar.
  • Layla & Wanderlog: AI travel agents. Layla handles discovery/booking; Wanderlog optimizes daily routes and logistics while traveling.
  • Magic School: Generates personalized tutoring content and educational activities.