Blog / AI Tooling
MY AI ToolKIT


4 Minute Read
Woody Kim
Woody Kim
January 7, 2026
Artificial Intelligence has come a long way since I was at the MIT AI Accelerator back in 2022. In just 3 years we have seen
Artificial Intelligence has come a long way since I was at the MIT AI Accelerator back in 2022. In just 3 years, we have gone from who has the biggest reasoning model to who has the best AI agents.
In 2026, I am back at MIT studying Deep Neural Networks and it is fascinating to think what was considered not achievable in decades have been accomplished in under 4 years.
What is the paradigm shift?
The cost to run GPT‑3.5-level systems dropped more than 280× between late 2022 and late 2024, driven by more efficient models, better inference, and hardware
improvements, making powerful AI accessible to many more users and products.


How does this affect Product Design
My Design AI Tooling Stack
At a fundamental level, I observe two trends in how companies are approaching the AI wave.
Agentic AI Pioneers: The first group is embracing multiple forms of tooling. These individuals, who are not afraid to burn tokens for ideation, are committed to rapid shipping, rapid debugging, and consistent production. Observability and Explainability play a significant role in this strategy, as issues may arise during AI prototyping and coding.
Legacy Adopters: The latter are operating on a 2023 understanding of AI. The most common use of AI I see daily is supercharging ChatGPT. While this is still an improvement to productivity, this can be greatly increased with slight tweaks to the tooling stack that I will share below. AI prototyping and coding.
My AI Tooling Overview
| Recommendation | ||||
|---|---|---|---|---|
| Claude Code by Anthropic | Using | Live Prototype | Paid | One of the main choices for vibe coding alongside Cursor. |
| Cursor | Using | Live Prototype | Paid | Preferred environment for vibe coding over Claude Code. |
| Figma Make | Using | Prototyping | Paid | Great for quick iterations; build patterns, refactor with auto layout, and clean up odd div structures. |
| HotJar | Not Using | Analytics | Free | User-friendly analytics with feature overlap with PostHog; good as an entry analytics tool. |
| Lovable | Not Using | Prototyping | Free / Paid | Produces a lot of low-craft output; heavy on gradients that feel flashy without adding functionality. |
| Magic Patterns | Using | Prototyping | Paid | Optimized for very fast prototypes; often used as a starting point before refining in Figma. |
| NotebookLM | Using | Research | Free | Condenses multi-source research into a single page, reducing the need for long-form video watching. |
| Perplexity | Using | Research | Paid | Excellent for research, formulation, and strong context gathering across mixed LLMs. |
| PostHog | Using | Analytics | Free | Developer-friendly analytics platform suitable for product teams. |
What is the paradigm shift?
The cost to run GPT‑3.5-level systems dropped more than 280× between late 2022 and late 2024, driven by more efficient models, better inference, and hardware
improvements, making powerful AI accessible to many more users and products.


How does this affect Product Design
My Design AI Tooling Stack
At a fundamental level, I observe two trends in how companies are approaching the AI wave.
Agentic AI Pioneers: The first group is embracing multiple forms of tooling. These individuals, who are not afraid to burn tokens for ideation, are committed to rapid shipping, rapid debugging, and consistent production. Observability and Explainability play a significant role in this strategy, as issues may arise during AI prototyping and coding.
Legacy Adopters: The latter are operating on a 2023 understanding of AI. The most common use of AI I see daily is supercharging ChatGPT. While this is still an improvement to productivity, this can be greatly increased with slight tweaks to the tooling stack that I will share below. AI prototyping and coding.
My AI Tooling Overview
| Recommendation | ||||
|---|---|---|---|---|
| Claude Code by Anthropic | Using | Live Prototype | Paid | One of the main choices for vibe coding alongside Cursor. |
| Cursor | Using | Live Prototype | Paid | Preferred environment for vibe coding over Claude Code. |
| Figma Make | Using | Prototyping | Paid | Great for quick iterations; build patterns, refactor with auto layout, and clean up odd div structures. |
| HotJar | Not Using | Analytics | Free | User-friendly analytics with feature overlap with PostHog; good as an entry analytics tool. |
| Lovable | Not Using | Prototyping | Free / Paid | Produces a lot of low-craft output; heavy on gradients that feel flashy without adding functionality. |
| Magic Patterns | Using | Prototyping | Paid | Optimized for very fast prototypes; often used as a starting point before refining in Figma. |
| NotebookLM | Using | Research | Free | Condenses multi-source research into a single page, reducing the need for long-form video watching. |
| Perplexity | Using | Research | Paid | Excellent for research, formulation, and strong context gathering across mixed LLMs. |
| PostHog | Using | Analytics | Free | Developer-friendly analytics platform suitable for product teams. |
My AI Tooling Overview
| Recommendation | ||||
|---|---|---|---|---|
| Claude Code by Anthropic | Using | Live Prototype | Paid | One of the main choices for vibe coding alongside Cursor. |
| Cursor | Using | Live Prototype | Paid | Preferred environment for vibe coding over Claude Code. |
| Figma Make | Using | Prototyping | Paid | Great for quick iterations; build patterns, refactor with auto layout, and clean up odd div structures. |
| HotJar | Not Using | Analytics | Free | User-friendly analytics with feature overlap with PostHog; good as an entry analytics tool. |
| Lovable | Not Using | Prototyping | Free / Paid | Produces a lot of low-craft output; heavy on gradients that feel flashy without adding functionality. |
| Magic Patterns | Using | Prototyping | Paid | Optimized for very fast prototypes; often used as a starting point before refining in Figma. |
| NotebookLM | Using | Research | Free | Condenses multi-source research into a single page, reducing the need for long-form video watching. |
| Perplexity | Using | Research | Paid | Excellent for research, formulation, and strong context gathering across mixed LLMs. |
| PostHog | Using | Analytics | Free | Developer-friendly analytics platform suitable for product teams. |