AI Use Cases – When is AI Helpful
- Stretegic thinking partner
- Writing and communication
- Software engineering copilot
- Product design partner
- Debugging human systems
- Prompt engineering for AI systems
- Scenario simulation
- Rapid knowledge acquisition
Strategic Thinking Partner
A strategy is a belief. For instance, you believe that a local-first approach will lead to a better app. But here's the thing, you don't know. You have a sense and maybe you have some experience building that approach, but you haven't seen every situation or even a lot of the situations.
AI helps think through complex problems, it surfaces second-order consequences, acts as a sparring partner and helps refine frameworks and language.
Examples
- "then what?" analysis of AI strategies of the big players
- Centralization vs decentralization architecture tradeoffs and decisions
- MCP vs APIs conceptual anslysis – similarities and differences
- Comparison of the dot-com and AI bubble for investment analysis
- Startup idea exploration
- Market intelligence
LLMs are really strong at pattern synthesis, generating alternative hypotheses and scenario modeling. AI can be a cognitive amplifier.
Writing and Communication
LLMs can quickly generate variations of conveying ideas as well as writing in different tones. It also does a great job identifying inconsistencies in writing which enables clearer writing.
Examples
- LinkedIn posts
- Marketing messaging
- Domain name ideas
- Slogan rewriting
- Landing page copy
- Storytelling formts (Enterprise, sci-fi, ect.)
- Proposal framing
- Positioning
LLMs are really good at language iteration. Utilizing AI for this use case is probably 10x faster than doing it manually.
Sofware Engineering Copilot
Software engineering includes system and architecture design, coding, testing, code and data generation. It compreses documentation searching, StackOverflow, and architectural reasoning.
LLMs are good at pattern recognition across messy inputs.
Examples
- P2P chat architecture
- Deterministic LLM execution runtime
- Event systems
- Database system comparisons
- MCP runtime design
- Converting from one framework to another
- Writing coding prompts
- Debugging code behavior
- Parse logs for troubleshooting
Product Design Partner
Product and system design requires understanding tradeoffs, different options, and thinking through how people use system in the real world.
LLMs ability to generate design alternatives quickly enables better design choices.
Debugging Human Systems
Social systems and incentives are a part of the problem space and LLMs are good at modeling them. Being able to incorporate these models into designing solutions just makes for better systems.
Prompt Engineering for Other AI Systems
Context management is a key lever to improve the determinism of the resulting generated code. LLMs are uniquely suited to help design instructions for other LLMs.
Scenario Simulation
Being able to think through alternative futures is a super power. LLMs simulate possible worlds well to imagine:
- The future of AI infrastructure
- A decentralized vs centralized runtime
- Econmis scenarios
It also enables analzying failur scenarios to design mitigations and SOPs.
Rapid Knowledge Acquisition
AI speeds up learning.
Across everything I've done, AI has played four primary roles:
| Role | Descriptoin |
|---|---|
| Thinking Partner | expand and challenge ideas |
| Engineering Copilot | accelerate building software |
| Research Assitant | compress information gathering |
| Design Partner | explore product and system design |
Failure Modes
- Bad apple – one reference to an old pattern can contribute to the LLM to reimplementing a bad or old pattern