From Prompts to Code: We are already learning a new way to speak with AI

Over the past few years, we have devoted time to learning how to write better prompts:

How to provide context to artificial intelligence.
How to explain exactly what we want.
How to define constraints.
How to tell it which sources to use.
How to ask for a specific format.
How to assign it a role.

We even created a new discipline around this skill: prompt engineering.
Yet recently, something interesting has been unfolding right before our eyes.
We need to write less and less.
Today, we use symbols such as @, #, or /, select agents, mention files, press buttons, open links, or use identifiers to trigger capabilities that we previously would have had to describe through much longer instructions.
In this new edition of Energy 4.0, I want to analyze this transition: how we are moving from drafting extensive instructions to interacting through references and identifiers. It is not that complexity has vanished, but rather that it has shifted from the user to the infrastructure. Prompts still exist, but they now reside within the invisible background architecture.
We are witnessing the shift from prompt engineering to context engineering.

1. From Explaining Everything to Simply Pointing

The first stage of generative AI popularized natural language as the universal computing interface: there was no longer a need to know Python or SQL to formulate a complex query. We could write:

“Analyze the behavior of the electricity market over the past week, compare spot prices with our budget, identify the main variances, explain their root causes, and prepare an executive summary highlighting key risks and opportunities.”

The model answered, but the entire cognitive and assembly burden fell on the operator. The user had to specify:

the objective;
the context;
the information;
the methodology;
the constraints;
and the format.

The architecture was essentially:

User → instructions + context → model → response

As models matured with RAG (Retrieval-Augmented Generation) architectures and tool/function calling capabilities, systems learned to retrieve context autonomously. Interaction stopped demanding the transcription of the problem and turned into a simple act of pointing.
That is where the next stage begins.

2. We Are Already Using Code to Speak with AI

This shift might seem futuristic until we examine how we use many artificial intelligence tools today. Various forms of reference-based interaction are already emerging.

@ — Invoking agents and tools:

Instead of dictating an extensive role to the model (“act as a quantitative analyst with access to the SCADA system”), the at-sign invokes the specific agent or microservice that already contains the predefined configuration.

# — Referencing data and repositories

Replaces the rudimentary act of copying and pasting tables or reports. By typing #DispatchReport_W34, the system knows which vector data or indexed document to inject into the context window.

/ — Executing workflows (Slash commands)

Commands like /analyze_variances encapsulate complex processing sequences, visualization parameters, and pre-programmed regulatory constraints.

Deep links (Deep Links and QR Codes)

A QR code on a power transformer in a substation is not merely a web link: it is a physical contextual anchor. Scanning it eliminates the need to tell the AI the brand, model, dissolved gas analysis (DGA) history, or last maintenance date; the identifier loads the asset’s complete ontology immediately.

3. The Prompt Did Not Disappear: It Became Infrastructure

What we are witnessing is not a simplification of the task, but a paradigm shift: from explicit context to dynamic, retrievable context.
In the energy sector, a dispatch analysis or portfolio review typically follows a predictable lifecycle:

Phase 1 (Artisanal Prompt): An analyst manually writes 300 words describing which variables to cross-reference.
Phase 2 (Template): After dozens of repetitions, the prompt is saved as a fixed template.
Phase 3 (Orchestrated Pipeline): The process is automated into a pipeline: prices → demand → renewable generation → fuel sources → variances → risk alerts → report.
Phase 4 (Unique Identifier): The user simply types /MarketReview or clicks a button.

Complexity was not eliminated; it was packaged into the underlying software layer.

4. Fewer Words Demand Greater Governance and Observability

This evolution presents a critical challenge: the less visible the instructions, the greater the need for internal traceability.

An explicit 500-word prompt leaves a clear audit log: we know exactly what the user requested and under what criteria. Conversely, a simple command like /MarketReview or an asset scan delegates orchestration to automated agents and function calls.

If visible interaction is minimal, the infrastructure must ensure:

Context Traceability (Data Provenance): Knowing precisely which data sources, model versions, and tools were consulted to generate the output.
Context Poisoning Mitigation: Preventing outdated data or anomalous sensor readings from corrupting agent reasoning undetected by the user.
Access Control and Security: Ensuring shortcuts respect operator authorization levels over operational technology (OT) infrastructure.

5. Catalog of Functional Shortcuts: Slash Commands in Action

This need to condense instructions is already evident in daily user workflows through functional shortcuts. Instead of drafting a paragraph detailing tone, audience, or technical depth, users turn to short commands (often preconfigured in Custom Instructions or system memory) to trigger immediate operational modes:

1. Marketing, Sales, and Conversion

/adcreative — Designs high-impact ad concepts with visual hooks, persuasive headlines, and CTAs.
/adspremium — Structures premium ad campaigns with high-value positioning and targeted segmentation.
/timesquaread — Designs large-format ads for urban digital screens and DOOH displays.
/ugcmodel — Creates guidelines and conversion-focused scripts for user-generated content.
/funnelbuilder — Builds digital sales funnels step by step.
/salesmentor — Develops commercial strategies to qualify leads and close deals.
/offerbuilder — Constructs high-converting offers with value propositions, bonuses, guarantees, and reduced friction.
/objectionkiller — Identifies and counters customer objections with structured reasoning.
/nichefinder — Discovers profitable, high-opportunity market niches.
/ctaexpert — Crafts calls to action tailored for maximum conversion.
/brandstory — Builds commercial and emotional brand narratives.
/audienceavatar — Builds detailed buyer personas and ideal audience profiles.
/launchplanner — Outlines end-to-end launch schedules for products or services.
/copywriter — Produces persuasive copy using frameworks like AIDA, PAS, and others.
/conversion — Optimizes content, offers, and proposals to boost conversion rates.

2. Technical Visualization, Design, and Photography

/storefront — Designs commercial facades, window displays, and exterior brand identity.
/xray — Visually reveals the hidden internal structure of an object.
/360view — Generates panoramic or immersive views of spaces and products.
/explodedview — Separates an object’s components to show internal structure and assembly.
/droneview — Creates top-down aerial perspectives to contextualize spaces and infrastructure.
/anatomy — Visually deconstructs an element and catalogs its component parts.
/cutaway — Produces cross-section diagrams displaying internal mechanics.
/skinpro — Enhances skin texture, micro-detail, and realism in portraits.
/deblur — Restores sharpness and corrects out-of-focus areas.
/removepeoplebgr — Removes background individuals from images.
/removebg — Completely removes image backgrounds.
/jawline — Sculpts and refines facial contours.
/3dposter — Creates 3D posters featuring volumetric typography and dramatic lighting.
/pantone — Defines color palettes using standard Pantone references and nomenclature.
/3dfloating — Produces zero-gravity aesthetic compositions with suspended objects.
/doubleexposure — Blends two scenes or concepts using double-exposure techniques.
/lifestyleshot — Positions a product inside realistic, everyday environments.
/3dsoft — Converts products into clean 3D renders with soft lighting and minimalist backgrounds.
/magazinecover — Turns images into professional magazine covers.
/visionnocturna — Applies realistic night-vision or thermal infrared looks.
/fotoproducto — Converts everyday photos into commercial studio product shots.
/esquema — Creates technical schematics complete with components, links, and callouts.
/dibujotecnico — Generates engineering blueprints with accurate proportions and dimensions.
/termico — Illustrates objects or spaces using thermal imaging and heatmaps.
/transparencia — Renders outer shells semi-transparent to reveal internal mechanics.
/malla3d — Converts 3D objects into wireframe structural representations.

3. Content, Social Media, and SEO

/viralhook — Crafts opening hooks designed to stop the scroll immediately.
/headlinegenius — Generates compelling headlines optimized for high click-through rates (CTR).
/contentrecycler — Repurposes a single concept across multiple formats and channels.
/contentoptimizer — Refines existing copy for search intent, clarity, and retention.
/carouselcreator — Designs slide-by-slide storytelling frameworks for carousels.
/reelsmaster — Scripts Reels, TikToks, and Shorts optimized for viewer retention.
/scriptwriter — Writes structured, dynamic video scripts.
/socialmediaai — Produces social posts tailored to each platform’s style.
/seo — Optimizes keywords, headers, layout, and search intent alignment.
/viral — Uncovers high-engagement angles and content concepts.

4. Writing, Editing, and Communication

/emailwizard — Drafts crisp, action-oriented executive emails.
/retroalimentacion — Evaluates work to highlight strengths and areas for improvement.
/editor — Refines clarity, flow, structure, and grammar.
/cambiartono — Rewrites text to match a specified tone of voice.
/adaptaraudiencia — Adjusts technical depth, register, and style for target audiences.

5. Strategy, Business, and Decision-Making

/ceo — Assesses scenarios through an executive lens focused on ROI, capital allocation, and risk.
/strategy — Develops long-range strategic plans with distinct phases, milestones, and risk controls.
/trendpredictor — Spots emerging market trends and projects forward trajectories.
/competitorroast — Pinpoints competitor flaws, operational gaps, and vulnerabilities.
/unpopularopinion — Builds rigorous counterarguments challenging prevailing consensus.
/moneyideas — Surfaces new business concepts and monetization paths.
/negotiationcoach — Simulates negotiations and recommends tactical moves.
/decisionhelper — Weighs strategic paths using weighted evaluation matrices.
/consultor — Analyzes business problems and delivers structured consulting recommendations.
/foda — Builds comprehensive SWOT analyses (Strengths, Weaknesses, Opportunities, Threats).
/proscontras — Compiles objective comparisons of advantages and drawbacks.
/compensaciones — Clarifies core trade-offs: what is gained versus sacrificed per option.
/primerosprincipios — Breaks problems down to fundamental first principles.
/benchmark — Evaluates performance against industry benchmarks and best practices.
/puntuacion — Scores alternatives against explicit, predefined criteria.

6. Analysis and Diagnostics

/deepthink — Executes rigorous, structured reasoning through technical problems.
/problemsolver — Deconstructs complex challenges to engineer step-by-step solutions.
/rootcause — Conducts root-cause diagnostics using frameworks like the 5 Whys.
/analista — Evaluates scenarios using data, formal logic, and systematic deduction.
/sistemas — Maps relationships, feedback loops, and dependencies across system components.
/analisisdebrechas — Conducts gap analyses between current states and target objectives.
/buscadorpatrones — Uncovers recurring trends and structural patterns within complex datasets.
/insight — Extracts high-leverage takeaways and core conclusions.
/profundizar — Progressively unpacks a subject from fundamental concepts to advanced nuances.

7. Risk, Critique, and Validation

/critic — Stress-tests proposals to uncover hidden vulnerabilities, risks, and flawed assumptions.
/segundaopinion — Delivers an independent, alternative perspective on decisions.
/analisisderiesgos — Catalogs potential hazards, vulnerabilities, and adverse consequences.
/revisarsupuestos — Surfaced underlying premises to audit their factual validity.
/verificarhechos — Separates verified facts from working assumptions and unconfirmed claims.
/abogadodeldiablo — Builds the strongest possible counterargument against a proposed plan.
/contraargumento — Outlines direct, robust rebuttals to specific claims.
/equiporojo — Runs red-team adversarial tests to identify operational vulnerabilities.
/obstaculos — Identifies probable bottlenecks and roadblocks before execution.
/revisionfinal — Conducts quality assurance reviews to catch inconsistencies and omissions.

8. Opportunity and Optimization

/analisisdeoportunidades — Surfaces overlooked, high-upside strategic opportunities.
/80-20 — Applies the Pareto Principle to pinpoint the 20% of actions driving 80% of returns.
/simplificador — Strips away unnecessary operational complexity from processes and ideas.
/enfoque — Clarifies high-priority focus areas while filtering out distractions.
/priorizar — Ranks action items based on expected impact, urgency, and resource demands.
/datosaccion — Converts raw data and analytical findings into concrete operating steps.

9. Planning and Execution

/plandeaccion — Translates ideas and high-level goals into tactical execution steps.
/hojaderuta — Maps out sequential roadmaps from current positioning to end milestones.
/hitos — Breaks overarching targets into distinct, trackable milestones.
/ejecucion — Converts abstract strategy into concrete operational plans.
/planalternativo — Develops contingency plans (Plan B) for unexpected disruptions.
/seguimiento — Converts goals into measurable accountability metrics and tracking cadences.
/antiprocrastinacion — Breaks intimidating tasks into micro-actions to kickstart momentum.

10. Research and Information Gathering

/research — Performs comprehensive desk research cross-referencing evidence and sources.
/buscadorfuentes — Curates top-tier primary and secondary sources to validate claims.
/creadordepreguntas — Formulates pivotal exploratory questions to guide deeper research.
/entrevistador — Uses progressive discovery questioning to diagnose issues before prescribing answers.

11. Productivity, Data, and Automation

/frameworkify — Structures abstract ideas into clean, repeatable frameworks.
/examplewizard — Produces concrete case examples to clarify conceptual explanations.
/datainterpreter — Analyzes complex data sets and extracts executive-level insights.
/automateit — Identifies automation candidates and maps digital workflows.
/productivityboost — Identifies workflow bottlenecks to optimize daily operating rhythms.

12. Presentations and Slide Decks

/slidebyslide — Outlines presentations slide by slide.
/pitchdeck — Builds structured 10–12 slide executive pitch decks.
/vba — Writes VBA scripts to automate PowerPoint deck generation.
/marp — Generates Markdown slide presentations compatible with Marp.
/slideprompt — Crafts optimized prompts for slide tools like Gamma, Beautiful.ai, and Tome.
/slidetext — Condenses dense paragraphs into punchy, high-impact slide copy.
/speakerdeck — Drafts presenter talking points and contextual speaker notes.
/visualconcept — Suggests visual metaphors, diagrams, and figures for slides.

13. Learning and Training

/teacher — Breaks down concepts systematically through structured pedagogical methods.
/skillbuilder — Maps learning roadmaps for skill acquisition.
/curriculo — Organizes broad topics into comprehensive curricula.
/cuestionario — Tests comprehension using progressively challenging questions.
/tarjetas — Turns key concepts into study flashcards.
/memoria — Generates analogies, associative devices, and mnemonics to boost recall.
/practica — Builds targeted exercises to develop practical proficiency.
/coach — Uses Socratic questioning to guide users toward their own answers.
/mentor — Provides practical career perspectives based on domain experience.

14. Career and Professional Development

/jobinterviewer — Runs realistic mock interviews and provides targeted feedback.
/careercoach — Builds career advancement and transition strategies.
/resumewriter — Formats achievement-driven resumes tailored for ATS parsing.
/coverlettergen — Writes personalized cover letters aligned to specific openings.
/linkedinbooster — Optimizes LinkedIn headlines, About sections, experience, and positioning.

15. Personal Growth, Habits, Wellness, and Finance

/personaldeveloper — Designs self-discipline and personal growth programs.
/habittracker — Builds habit tracking schedules, metrics, and cadences.
/healthcoach — Delivers general guidance on healthy lifestyle habits and preventive routines.
/mealplanner — Builds structured weekly meal plans.
/workoutbuddy — Creates goal-specific fitness routines.
/travelplanner — Designs travel itineraries balancing time, transit logistics, and budgets.
/budgetplanner — Builds budgeting systems and expense-tracking routines.
/financetips — Explains core personal finance concepts across savings, debt, and investing.
/booksummary — Summarizes books highlighting actionable takeaways and core lessons.

16. Storytelling, Travel, and Narrative

/storytellerai — Structures information or personal experiences into compelling narrative arcs.
/psychowords — Integrates persuasive psychological language and power words.
/timeandlocation — Sets scenes with rich historical, geographic, and temporal details.
/memorytrail — Weaves chronological narratives from fragmented memories and experiences.
/localmoment — Captures local color, regional customs, and neighborhood atmosphere.
/letterfromthere — Writes personal first-person letters sent from travel destinations.
/passportmemory — Structures trips into travel journal entries capturing key reflections.

17. Formatting, Synthesis, and Meta-AI

/brief — Delivers ultra-concise answers, eliminating conversational filler.
/table — Formats findings into comparison tables or structured grids.
/eli5 — Explains complex concepts simply using everyday analogies (Explain Like I’m 5).
/human — Writes with a natural, human rhythm while avoiding formulaic AI patterns.
/expert — Answers with high technical rigor from the perspective of a senior specialist.
/promptengineer — Analyzes and refines prompts to maximize AI output quality.

6. Practical Guide: How to Configure Your Own Custom Slash Commands

To translate theory into an operational workflow within ChatGPT or similar interfaces, you do not need to code an application from scratch. You can leverage Custom Instructions, System Memories, or custom GPT configurations.

Here is a step-by-step setup guide:

Step 1: Design the macro instruction (The Hidden Prompt):Before creating a shortcut, define precisely what the model must do when triggered. Avoid ambiguity.

Example for /deepthink
When the user types /deepthink, adopt the role of a senior research scientist. Break your answer down into: 1) Core Hypothesis, 2) Step-by-step Deductive Reasoning, 3) Critical Variables, and 4) Validated Conclusion. Omit introductory conversational filler.

Step 2: Centralize commands in system memory:Add a clean definitions block directly into your Custom Instructions (under How would you like ChatGPT to respond?) or in the instructions field of your custom GPT using a dictionary structure:

RAPID ACCESS PROTOCOL (SLASH COMMANDS):
- /deepthink: Activates structured deductive reasoning across 4 phases (Hypothesis, Logic, Variables, Conclusion).
- /brief: Direct, condensed output. No conversational openings or polite sign-offs.
- /storefront: Models the visual design, facade, and architectural elements of a retail space.
- /adspremium: Structures high-ticket advertising copy using an exclusive tone and persuasive CTA.

Step 3: Establish formatting and fallback rules: Add a master instruction specifying how the model should handle unknown or mistyped commands:

If the user inputs a slash command (/) not present in the list above, infer its semantic intent or request clarification in fewer than 10 words.

Step 4: Test and refine the pipeline: Stress-test the setup by chaining commands (for example: #Q3_FinancialReport + /datainterpreter + /brief). Check whether the background context is retrieved cleanly or if injected context parameters need fine-tuning. Packing instructions into short commands reduces day-to-day operating friction and keeps output quality consistently professional, regardless of how brief the input message is.

6. Conclusion

We spent years learning how to write better prompts. Now, almost imperceptibly, we are learning to write less: an @, a #, a /, a link, a QR code.

It would be a mistake, however, to view this simply as simplification. The underlying task has not become simpler; rather, the complexity has receded from view, migrating from our visible prompt into the supporting infrastructure behind it.

While this delivers significant productivity gains, it brings a corresponding challenge: the less we write, the more trust we place in the system’s ability to interpret our intent accurately. That trust depends on having a resilient, observable infrastructure behind every shorthand code—one with clear end-to-end traceability for every action taken.

Be sure to bookmark this guide so you have these shortcut commands handy whenever you need them!

References

Context Engineering: Why Your AI Agent Needs a Database, Not a Prompt https://dev.to/monuminu/context-engineering-why-your-ai-agent-needs-a-database-not-a-prompt-1j79

Prompt Engineering is Dead. Long Live Context-as-Code https://builder.aws.com/content/3EaojDFlYgdiXzRqJ1e94xbkgc3/prompt-engineering-is-dead-long-live-context-as-code