DEV Community

Cover image for 100 Creative Thinking, AI Workflows & Advanced Prompting Lenses 🤖🧠
Probal Dhali
Probal Dhali

Posted on

100 Creative Thinking, AI Workflows & Advanced Prompting Lenses 🤖🧠

The first 400 lenses focused on writing, visual thinking, productivity, coding, research, and content creation.

This final collection moves one layer deeper:

How do we design better interactions with AI itself?

Instead of treating AI as a single chatbot, we can think of it as a configurable reasoning and workflow environment.

Question
   ↓
Context
   ↓
Lens
   ↓
Method
   ↓
Evaluation
   ↓
Iteration
   ↓
Result
Enter fullscreen mode Exit fullscreen mode

The goal isn't to make prompts unnecessarily complicated.

The goal is to make the task, constraints, perspective, and evaluation criteria explicit.

1. Creative Generation

The foundation is idea generation.

/brainstorm
/ideas10
/ideas50
/creative
/innovate
/outofthebox
/inspiration
/random
Enter fullscreen mode Exit fullscreen mode

A simple creative workflow:

Problem
  ↓
Generate
  ↓
Expand
  ↓
Combine
  ↓
Filter
  ↓
Evaluate
Enter fullscreen mode Exit fullscreen mode

For example:

/ideas50
Enter fullscreen mode Exit fullscreen mode

is useful when quantity matters.

But:

/bestoption
Enter fullscreen mode Exit fullscreen mode

is useful when the goal shifts from exploration → selection.

That distinction is important.

2. Remixing Existing Ideas

Not every useful idea needs to start from zero.

These lenses work with existing concepts:

/combine
/remix
/alternate
/variants
/options
/compareideas
Enter fullscreen mode Exit fullscreen mode

A powerful pattern is:

Existing Idea
     ↓
Decompose
     ↓
Change One Variable
     ↓
Generate Variant
     ↓
Compare
Enter fullscreen mode Exit fullscreen mode

For example:

Existing:
AI study assistant

Variant A:
AI study planner

Variant B:
AI exam simulator

Variant C:
AI knowledge graph

Variant D:
AI peer tutor
Enter fullscreen mode Exit fullscreen mode

Creativity often comes from recombination, not pure novelty.

3. Decision-Making

Once ideas exist, selection becomes the problem.

Useful lenses:

/bestoption
/decision
/options
/compareideas
Enter fullscreen mode Exit fullscreen mode

Instead of:

“Which one is best?”

a structured decision asks:

Criteria
 ↓
Weights
 ↓
Options
 ↓
Evidence
 ↓
Trade-offs
 ↓
Score
 ↓
Decision
Enter fullscreen mode Exit fullscreen mode

This reduces the temptation to choose the first attractive idea.

4. Advisor → Mentor → Coach

Different situations require different interaction styles:

/advisor
/mentor
/coach
/teacher
/tutor
Enter fullscreen mode Exit fullscreen mode

They can be thought of as different levels of guidance.

Advisor
   ↓
Recommendations

Mentor
   ↓
Experience + direction

Coach
   ↓
Questions + accountability

Teacher
   ↓
Explanation

Tutor
   ↓
Personalized practice
Enter fullscreen mode Exit fullscreen mode

The important point is that the interaction mode changes, not necessarily the underlying knowledge.

5. Expert Perspectives

Different disciplines frame problems differently.

This collection includes:

/professor
/scientist
/engineer
/doctor
/lawyer
/psychologist
/economist
/historian
/journalist
Enter fullscreen mode Exit fullscreen mode

These should be treated as perspective lenses, not claims of real-world professional credentials.

For example:

Same problem
      ↓
 ┌────┼────┐
 ↓    ↓    ↓
Engineer Economist Historian
 ↓    ↓    ↓
System Cost Context
Enter fullscreen mode Exit fullscreen mode

A multidisciplinary view can reveal assumptions that one perspective misses.

For medical or legal topics, the output should remain educational and should not be treated as a substitute for qualified professional advice.

6. Professional Roles

The next layer applies organizational perspectives:

/editor
/designer
/architect
/productmanager
/founder
/ceo
/investorview
/customer
Enter fullscreen mode Exit fullscreen mode

Imagine evaluating a product:

Engineer
→ Can we build it?

Product Manager
→ Should we build it?

Founder
→ Is the opportunity worth pursuing?

Investor
→ Is the business attractive?

Customer
→ Does it actually solve my problem?
Enter fullscreen mode Exit fullscreen mode

These perspectives can disagree.

And that disagreement can be valuable.

7. Beginner vs Advanced

Two useful control lenses:

/beginner
/advanced
Enter fullscreen mode Exit fullscreen mode

The same concept can be represented differently.

Beginner:
Simple language
Examples
Analogies
Few assumptions

Advanced:
Formal terminology
Edge cases
Constraints
Trade-offs
Implementation details
Enter fullscreen mode Exit fullscreen mode

A good AI workflow should adapt complexity to the audience.

8. Interactive Problem Solving

Some problems shouldn't be solved in one giant response.

Useful lenses:

/stepbystep
/interactive
/walkthrough
Enter fullscreen mode Exit fullscreen mode

Instead of:

Question → Final Answer
Enter fullscreen mode Exit fullscreen mode

you can use:

Question
 ↓
Clarify
 ↓
Analyze
 ↓
Confirm
 ↓
Execute
 ↓
Review
Enter fullscreen mode Exit fullscreen mode

This is particularly useful for complex technical tasks where requirements may be ambiguous.

9. Simulation

The /simulation lens can model hypothetical situations.

For example:

Current System
      ↓
Change Condition
      ↓
Simulate Outcome
      ↓
Identify Effects
      ↓
Evaluate
Enter fullscreen mode Exit fullscreen mode

Possible applications include:

  • system design
  • product decisions
  • interview practice
  • business scenarios
  • negotiation
  • project planning

But simulations should be clearly treated as scenarios, not predictions of reality.

10. Debate and Counterarguments

Good reasoning requires the ability to challenge an idea.

Useful lenses:

/debate
/counterargument
/critique
/validate
Enter fullscreen mode Exit fullscreen mode

A useful structure:

Claim
 ↓
Supporting Evidence
 ↓
Counterargument
 ↓
Response
 ↓
Remaining Weakness
 ↓
Conclusion
Enter fullscreen mode Exit fullscreen mode

This prevents an AI conversation from becoming an automatic agreement machine.

11. Assumptions and Constraints

Two of the most underrated lenses:

/assumptions
/constraints
Enter fullscreen mode Exit fullscreen mode

Every project contains assumptions.

For example:

“This API will scale.”
Enter fullscreen mode Exit fullscreen mode

Possible hidden assumptions:

Expected traffic
Infrastructure
Database performance
Network conditions
Caching
Concurrency
Failure handling
Enter fullscreen mode Exit fullscreen mode

Making assumptions visible creates opportunities to test them.

12. Trade-Off Analysis

Engineering is full of trade-offs.

/tradeoffs
Enter fullscreen mode Exit fullscreen mode

can frame decisions like:

Performance ↔ Maintainability

Cost ↔ Reliability

Speed ↔ Quality

Flexibility ↔ Simplicity

Security ↔ Convenience
Enter fullscreen mode Exit fullscreen mode

There is rarely a solution that maximizes every dimension simultaneously.

The important question becomes:

Which trade-off is acceptable for this specific context?

13. Future Thinking

Future-oriented lenses include:

/opportunities
/risksfuture
/trendanalysis
/forecastfuture
/signals
/emergingtech
/futureproof
Enter fullscreen mode Exit fullscreen mode

A useful structure:

Current State
     ↓
Signals
     ↓
Possible Changes
     ↓
Scenarios
     ↓
Risks
     ↓
Opportunities
Enter fullscreen mode Exit fullscreen mode

The goal isn't to pretend AI can predict the future with certainty.

The goal is to explore plausible scenarios.

14. Weak Signals

The /signals lens focuses on early indicators.

For example:

Small change
   ↓
Repeated occurrence
   ↓
Emerging pattern
   ↓
Potential trend
Enter fullscreen mode Exit fullscreen mode

This can be useful for:

  • technology
  • product strategy
  • research
  • market analysis

But weak signals are inherently uncertain.

They should be treated as hypotheses, not facts.

15. Automation

The next group focuses on turning repetitive work into workflows.

/automation
/workflow
/pipeline
/automationplan
Enter fullscreen mode Exit fullscreen mode

A basic automation pipeline:

TRIGGER
   ↓
INPUT
   ↓
PROCESS
   ↓
VALIDATION
   ↓
OUTPUT
   ↓
LOGGING
Enter fullscreen mode Exit fullscreen mode

The key is not:

“Automate everything.”

Instead:

Automate predictable, repeatable work where the cost of errors is understood.

16. Systems Thinking

Complex problems rarely exist in isolation.

/systemthinking
Enter fullscreen mode Exit fullscreen mode

encourages looking at:

Component
   ↓
Relationships
   ↓
Feedback loops
   ↓
Dependencies
   ↓
Emergent behavior
Enter fullscreen mode Exit fullscreen mode

For example:

Users
 ↕
Product
 ↕
Infrastructure
 ↕
Data
 ↕
Business
Enter fullscreen mode Exit fullscreen mode

Changing one part can affect the others.

17. Mental Models

Useful lenses:

/mentalmodels
/heuristics
/principles
Enter fullscreen mode Exit fullscreen mode

Examples of mental models include:

First principles
Opportunity cost
Second-order effects
Feedback loops
Incentives
Bottlenecks
Compounding
Enter fullscreen mode Exit fullscreen mode

The value comes from selecting a model appropriate to the problem.

18. Framework Selection

Instead of automatically applying one framework:

/frameworkcompare
Enter fullscreen mode Exit fullscreen mode

can compare alternatives.

For example:

Problem
 ↓
Candidate Frameworks
 ↓
Strengths
 ↓
Weaknesses
 ↓
Fit
 ↓
Selection
Enter fullscreen mode Exit fullscreen mode

The right framework depends on the problem.

19. Best Practices vs Context

Useful lenses:

/bestpractice
/mistakes
/dosanddonts
/quickwins
Enter fullscreen mode Exit fullscreen mode

But “best practice” should never mean:

“Always do X.”

A better formulation is:

Common practice
      +
Context
      +
Constraints
      ↓
Recommended approach
Enter fullscreen mode Exit fullscreen mode

A practice that works in one organization can fail in another.

20. Optimization

The optimization group:

/optimization
/efficiency
/quickwins
Enter fullscreen mode Exit fullscreen mode

should begin with measurement.

A useful engineering principle:

Measure
 ↓
Identify bottleneck
 ↓
Change
 ↓
Measure again
Enter fullscreen mode Exit fullscreen mode

Optimization without measurement can simply create complexity without measurable improvement.

21. AI Workflows

Now we reach the heart of this collection:

/aiworkflow
/workflow
/pipeline
Enter fullscreen mode Exit fullscreen mode

Instead of one huge prompt:

Prompt
 ↓
Answer
Enter fullscreen mode Exit fullscreen mode

we can construct:

Input
 ↓
Analysis
 ↓
Generation
 ↓
Verification
 ↓
Revision
 ↓
Final Output
Enter fullscreen mode Exit fullscreen mode

This is often more controllable because each stage has a defined purpose.

22. Prompt Auditing

The /promptaudit lens evaluates the prompt itself.

Possible dimensions:

Goal
Context
Constraints
Ambiguity
Output format
Evaluation criteria
Missing information
Enter fullscreen mode Exit fullscreen mode

For example:

Weak:

“Build me a good website.”

Better:

Goal
+ Audience
+ Features
+ Tech stack
+ Constraints
+ Acceptance criteria
+ Output format
Enter fullscreen mode Exit fullscreen mode

The important improvement isn't necessarily length.

It is specificity where specificity matters.

23. Prompt Libraries

The /promptlibrary lens organizes reusable prompts.

A useful structure:

/prompts
 ├── coding
 ├── research
 ├── writing
 ├── analysis
 ├── debugging
 ├── planning
 └── evaluation
Enter fullscreen mode Exit fullscreen mode

Each prompt can include:

Name
Purpose
Inputs
Constraints
Expected Output
Example
Version
Enter fullscreen mode Exit fullscreen mode

This turns ad-hoc prompting into a reusable system.

24. Prompt Chaining

The /promptchain lens connects multiple stages.

Example:

Prompt 1
Research
 ↓
Prompt 2
Extract evidence
 ↓
Prompt 3
Analyze
 ↓
Prompt 4
Critique
 ↓
Prompt 5
Generate final output
Enter fullscreen mode Exit fullscreen mode

This can be more reliable than asking for everything simultaneously.

25. Meta-Prompting

The /metaprompt lens asks AI to create a prompt for another task.

Conceptually:

User Goal
    ↓
Meta-Prompt
    ↓
Task Prompt
    ↓
AI Task
    ↓
Output
Enter fullscreen mode Exit fullscreen mode

For example:

Goal:
Create a technical tutorial.

Meta-prompt:
“Design the optimal prompt structure for generating
a technically accurate tutorial for intermediate developers.”
Enter fullscreen mode Exit fullscreen mode

The meta-layer is useful when the task itself is complex.

26. System Prompts

The /systemprompt lens focuses on defining persistent behavior.

A useful conceptual structure:

ROLE
 ↓
OBJECTIVE
 ↓
RULES
 ↓
CONSTRAINTS
 ↓
TOOLS
 ↓
OUTPUT FORMAT
 ↓
EVALUATION
Enter fullscreen mode Exit fullscreen mode

This is closer to behavior specification than ordinary prompting.

27. Personas and Roles

Two related lenses:

/persona
/role
Enter fullscreen mode Exit fullscreen mode

A persona might specify:

Audience
Tone
Expertise
Communication style
Priorities
Enter fullscreen mode Exit fullscreen mode

But personas should not be confused with genuine professional credentials.

“Act as a senior architect” is a useful perspective instruction.

It does not make the AI a real architect.

28. Multi-Expert Reasoning

The /multiexpert lens can intentionally introduce multiple perspectives.

Example:

             PROBLEM
                │
      ┌─────────┼─────────┐
      ↓         ↓         ↓
  Engineer   Researcher  Product
      ↓         ↓         ↓
      └─────────┼─────────┘
                ↓
           Comparison
                ↓
             Decision
Enter fullscreen mode Exit fullscreen mode

This can expose disagreements that a single perspective might miss.

29. Consensus vs Disagreement

Useful lenses:

/consensus
/disagreement
Enter fullscreen mode Exit fullscreen mode

Instead of forcing agreement:

Expert A → Option X
Expert B → Option Y
Expert C → Option X
Enter fullscreen mode Exit fullscreen mode

the system can explicitly identify:

Areas of agreement
Areas of disagreement
Evidence differences
Assumption differences
Enter fullscreen mode Exit fullscreen mode

Disagreement can be information.

30. Evidence

The /evidence lens encourages claims to be connected to supporting information.

A useful structure:

CLAIM
 ↓
EVIDENCE
 ↓
SOURCE
 ↓
RELEVANCE
 ↓
LIMITATION
Enter fullscreen mode Exit fullscreen mode

This is especially important for:

  • research
  • technical writing
  • statistics
  • health information
  • legal topics
  • current events

AI-generated claims should still be independently verified when accuracy matters.

31. Confidence and Uncertainty

Two complementary lenses:

/confidence
/uncertainty
Enter fullscreen mode Exit fullscreen mode

A useful answer separates:

Known
Likely
Possible
Unknown
Enter fullscreen mode Exit fullscreen mode

Instead of presenting every statement with the same level of certainty.

This is especially important when evidence is incomplete.

32. Limitations

The /limitations lens forces the question:

“Where could this approach fail?”

A useful structure:

Method
 ↓
Strengths
 ↓
Weaknesses
 ↓
Failure Conditions
 ↓
Uncertainty
 ↓
Mitigation
Enter fullscreen mode Exit fullscreen mode

This is one of the simplest ways to make AI-generated analysis more intellectually honest.

33. Comprehensive Assessment

The /assess lens combines multiple dimensions:

Correctness
Feasibility
Cost
Risk
Maintainability
Evidence
Constraints
Alternatives
Enter fullscreen mode Exit fullscreen mode

Instead of:

“Is this good?”

we ask:

“Good according to which criteria?”

That is a much better question.

34. Scorecards

The /scorecard lens makes qualitative decisions more explicit.

Example:

Criterion          Weight
──────────────────────────
Cost                 20%
Performance          25%
Security             25%
Maintainability      20%
Scalability          10%
Enter fullscreen mode Exit fullscreen mode

Then:

Option A → Score
Option B → Score
Option C → Score
Enter fullscreen mode Exit fullscreen mode

The numbers don't magically make the decision objective.

They simply make the evaluation criteria visible.

35. Ranking and Comparison Matrices

Useful lenses:

/ranking
/matrixcompare
Enter fullscreen mode Exit fullscreen mode

For example:

              Cost  Speed  Security  Scale
Option A       8      7       9        6
Option B       6      9       7        9
Option C       9      6       8        7
Enter fullscreen mode Exit fullscreen mode

The matrix makes trade-offs easier to inspect.

36. Final Answer Mode

The /finalanswer lens is intentionally simple:

/finalanswer
Enter fullscreen mode Exit fullscreen mode

It means:

Produce the requested final output without unnecessary process narration.

This is useful when the analysis has already happened and only the deliverable is needed.

37. High-Level Reasoning

The /thinking lens needs an important distinction.

It should not mean exposing private chain-of-thought.

Instead, it can request a concise reasoning summary:

Problem
 ↓
Key assumptions
 ↓
Relevant evidence
 ↓
Decision criteria
 ↓
Conclusion
Enter fullscreen mode Exit fullscreen mode

For example:

Reasoning summary:
- Constraint A eliminates option B.
- Requirement C favors option A.
- Main uncertainty is D.
- Therefore option A is currently preferred.
Enter fullscreen mode Exit fullscreen mode

That gives useful transparency without requiring hidden internal reasoning.

38. Execution Mode

The /executemode lens focuses on producing the requested artifact.

For example:

Goal:
Create project structure.

EXECUTE
 ↓
Files
 ↓
Implementation
 ↓
Tests
 ↓
Validation
Enter fullscreen mode Exit fullscreen mode

This is particularly useful when the planning stage is already complete.

39. Master Prompt

The /masterprompt lens combines multiple techniques.

Conceptually:

MASTER PROMPT
      │
      ├── Context
      ├── Goal
      ├── Role
      ├── Constraints
      ├── Process
      ├── Evidence
      ├── Evaluation
      └── Output Format
Enter fullscreen mode Exit fullscreen mode

But bigger is not automatically better.

A master prompt should remain understandable and maintainable.

40. The All-in-One Lens

Finally:

/allinone
Enter fullscreen mode Exit fullscreen mode

can combine the most relevant structures automatically.

A conceptual pipeline:

                 USER GOAL
                     ↓
              CONTEXT ANALYSIS
                     ↓
              TASK CLASSIFICATION
                     ↓
            SELECT BEST LENSES
                     ↓
          ┌──────────┼──────────┐
          ↓          ↓          ↓
       REASON      CREATE     VERIFY
          └──────────┼──────────┘
                     ↓
                  REVIEW
                     ↓
                 FINALIZE
Enter fullscreen mode Exit fullscreen mode

The important word is relevant.

Using every possible technique on every problem would create unnecessary complexity.

The Complete 100-Lens Map

# Shortcut Purpose
401 /brainstorm Generate many ideas
402 /ideas10 Generate 10 ideas
403 /ideas50 Generate 50 ideas
404 /creative Creative thinking
405 /innovate Innovative solutions
406 /outofthebox Unconventional ideas
407 /inspiration Creative inspiration
408 /random Random idea generation
409 /combine Combine concepts
410 /remix Remix existing ideas
411 /alternate Alternative approaches
412 /variants Generate variants
413 /options List options
414 /compareideas Compare ideas
415 /bestoption Select strongest option
416 /decision Decision support
417 /advisor Advisor perspective
418 /mentor Mentorship mode
419 /coach Coaching mode
420 /teacher Step-by-step teaching
421 /tutor Personalized tutoring
422 /professor University-level perspective
423 /scientist Scientific perspective
424 /engineer Engineering perspective
425 /doctor Medical educational perspective
426 /lawyer Legal educational perspective
427 /psychologist Psychology perspective
428 /economist Economic analysis
429 /historian Historical perspective
430 /journalist Investigative perspective
431 /editor Editorial perspective
432 /designer Design-thinking perspective
433 /architect Architecture mindset
434 /productmanager Product management view
435 /founder Founder perspective
436 /ceo Executive decision perspective
437 /investorview Investor evaluation
438 /customer Customer perspective
439 /beginner Beginner explanation
440 /advanced Advanced explanation
441 /stepbystep Sequential instructions
442 /interactive Interactive problem solving
443 /walkthrough Detailed walkthrough
444 /simulation Scenario simulation
445 /negotiation Negotiation simulation
446 /debate Opposing viewpoints
447 /counterargument Counterarguments
448 /critique Critical evaluation
449 /improveidea Strengthen an idea
450 /validate Validate an idea
451 /assumptions Identify assumptions
452 /constraints Identify constraints
453 /tradeoffs Explain trade-offs
454 /opportunities Find opportunities
455 /risksfuture Future risk analysis
456 /trendanalysis Analyze trends
457 /forecastfuture Explore possible futures
458 /signals Detect weak signals
459 /emergingtech Explore emerging technology
460 /futureproof Future-ready planning
461 /automation Find automation opportunities
462 /workflow Design workflows
463 /pipeline Create process pipelines
464 /systemthinking Systems analysis
465 /mentalmodels Apply mental models
466 /heuristics Decision heuristics
467 /principles Identify core principles
468 /frameworkcompare Compare frameworks
469 /bestpractice Industry practices
470 /mistakes Identify common mistakes
471 /dosanddonts Do's and Don'ts
472 /quickwins Fast improvements
473 /optimization Optimize performance
474 /efficiency Improve efficiency
475 /automationplan Automation roadmap
476 /aiworkflow AI-assisted workflow
477 /promptaudit Audit prompts
478 /promptlibrary Organize prompts
479 /promptchain Chain prompts
480 /metaprompt Generate prompts
481 /systemprompt Draft system prompts
482 /persona Create personas
483 /role Define a role
484 /simulateexpert Expert-perspective simulation
485 /multiexpert Multiple perspectives
486 /consensus Find agreement
487 /disagreement Identify disagreement
488 /evidence Support claims with evidence
489 /confidence Estimate confidence
490 /uncertainty Explain uncertainty
491 /limitations Identify limitations
492 /assess Comprehensive assessment
493 /scorecard Create scoring framework
494 /ranking Rank options
495 /matrixcompare Comparison matrix
496 /finalanswer Output only final answer
497 /thinking High-level reasoning summary
498 /executemode Focus on execution
499 /masterprompt Combine prompting techniques
500 /allinone Combine suitable structures

The 500-Lens Architecture

Now the bigger picture becomes interesting.

The entire collection can be represented as five layers:

                 AI WORKBENCH
                      │
     ┌────────────────┼────────────────┐
     ↓                ↓                ↓
VISUAL THINKING   CONTENT CREATION   IMAGE GENERATION
  001–100            101–200            201–300
     │                │                │
     └────────────────┼────────────────┘
                      ↓
              PRODUCTIVITY + CODE
                   301–400
                      │
                      ↓
          CREATIVE + AI WORKFLOWS
                   401–500
Enter fullscreen mode Exit fullscreen mode

Together:

001–100   → THINK
101–200   → CREATE
201–300   → VISUALIZE
301–400   → BUILD
401–500   → STRATEGIZE
Enter fullscreen mode Exit fullscreen mode

That's the real concept behind the collection.

From 500 Commands to One System

A sophisticated AI workflow might look like:

               USER GOAL
                   ↓
             /brainstorm
                   ↓
              /options
                   ↓
            /compareideas
                   ↓
             /constraints
                   ↓
              /tradeoffs
                   ↓
             /systemthinking
                   ↓
              /aiworkflow
                   ↓
              /promptchain
                   ↓
                /audit
                   ↓
              /finalanswer
Enter fullscreen mode Exit fullscreen mode

Each lens performs a different function.

The result is not necessarily a “better prompt.”

It is a better workflow.

The Most Important Lesson

After building a collection of 500 lenses, one principle stands out:

The quality of an AI interaction depends heavily on whether the task has been properly framed.

A useful conceptual model is:

AI Output Quality
        ≈
Context
× Task Definition
× Constraints
× Relevant Perspective
× Verification
Enter fullscreen mode Exit fullscreen mode

This is not a scientific equation or guaranteed quality formula—it's a useful design model.

And it highlights something important:

Adding more words isn't the same as adding better instructions.

The Future of Prompting?

Maybe the future isn't:

ONE GIANT PROMPT
Enter fullscreen mode Exit fullscreen mode

Maybe it looks more like:

                 AI SYSTEM
                     │
        ┌────────────┼────────────┐
        ↓            ↓            ↓
      LENS         WORKFLOW     EVALUATOR
        │            │            │
        ↓            ↓            ↓
   Perspective    Process      Verification
        └────────────┼────────────┘
                     ↓
                  OUTPUT
Enter fullscreen mode Exit fullscreen mode

Instead of memorizing hundreds of commands, an AI assistant could eventually infer:

“This is a research problem, so I need evidence, uncertainty, assumptions, and evaluation.”

Or:

“This is a debugging task, so I need reproduction steps, constraints, logs, hypotheses, and regression testing.”

That is a much more interesting direction than simply making prompts longer.

Final Question

If you could add one #501 lens to this collection, what would it be?

Mine would probably be:

/verify

Because generating an answer is only half the problem.

Knowing whether the answer deserves to be trusted is the other half.

AI #PromptEngineering #ArtificialIntelligence #AITools #Productivity #Programming #DeveloperTools #Research #CreativeThinking #ChatGPT #DevCommunity

Top comments (0)