Coding with AI

Prompt Engineering for AI

Last updated 2026-09-19

What's new

2026-09-19

You can save on usage limits by using smaller, cheaper AI models (like Sonnet or Haiku) for simple tasks, and reserving the more expensive model (Fable) for complex work.
Sub-agents (smaller AI models that Fable can assign tasks to) can handle tasks like reading files or making edits, saving your usage limit for more important work.
You can create custom sub-agents with specific instructions and reasoning levels to further optimize costs and efficiency.
By setting up rules in your Claude.md file, you can automate which tasks are delegated to which sub-agents, saving you time and tokens.

2026-09-13

Claude's Fable (an AI tool) and Grockbot (a cloud-based computer) can now handle most of the work in running an agency, from finding leads to sending emails, reducing the need for hiring people.
Fable is used for "thinking" tasks like deciding who to target and writing proposals, while Grockbot handles "doing" tasks like sending emails and logging into software.
Using APIs (a way for software to communicate directly) can make tasks cheaper and more effective, but Grockbot is useful for tasks where no API exists.
Instead of selling "AI automations," focus on selling the results, like getting 10 qualified calls on a client's calendar each month, to make your offer more appealing.

2026-09-07

Claude Fable 5.1 is a new AI model that can work on tasks independently, using multiple tools and programs, like a smart assistant (called an agent) to achieve goals you set for it.
It can create detailed 3D models and designs, like a fully furnished apartment, based on a simple floor plan, showing strong spatial understanding.
The model also demonstrated advanced physics and lighting understanding by creating a ray tracing simulation of shapes floating in an ocean with adjustable settings, all coded from scratch.

2026-09-04

Claude Fable 5.1 (an AI tool) is a big upgrade from Fable 5, especially for scientific research, with double the performance in some areas.
It's more efficient, saving 25-40% on tasks, making it cheaper to use than the previous version.
Fable 5.1 has better filtering, answering more questions without over-censoring like Fable 5 did.
It outperformed other AI models in tests, like creating detailed 3D roller coasters and finding bugs in code.

2026-08-25

**Codex** (an AI tool you install on your computer) helps businesses automate tasks like sales, marketing, and operations to save time and increase profits.
Learn to turn sales calls into **custom proposals with e-signatures and checkout** (ready-to-send documents that clients can sign and pay for online).
**Repurpose content** (reuse one video or post to create multiple social media posts, newsletters, or short clips) to save time and reach more people.
Build an **after-hours voice agent** (an AI that answers calls and handles leads automatically) to catch missed opportunities and improve customer service.

2026-08-22

**Cut absolute rule words** like "always," "never," and "must" from your AI prompts, as they force the AI to follow instructions too literally, often sacrificing quality (AI is a computer program that generates responses based on input).
**Replace vague words** like "detailed" or "professional" with clear definitions or examples, as the AI's understanding of these terms may differ from yours.
**Remove phrases** like "think harder" or "step by step," as they no longer help and can distract the AI (AI is a computer program that generates responses based on input).
**Use if-then statements** to give the AI flexibility, like "If a section has no real risks, then write 'not found' instead of inventing one."

2026-08-19

Claude (an AI tool) can act like an AI employee if set up with a workspace (a 'repo' or project folder), memory (context about the project), and a clear assignment (a 'ticket').
It can also have 'eyes' to inspect and test the product, review changes, and follow a schedule (called 'routines') for recurring tasks.
Permissions can be set to limit Claude's actions, ensuring it doesn't make risky decisions without human approval.
The goal is to create an AI-native company, where Claude operates as an always-on, productive team member.

2026-08-07

To stand out in AI, focus on showing real results (like time saved or leads gained) rather than just showcasing what you've built, as this helps prove your skills to potential clients or employers.
Don't rely too heavily on specific tools (like n8n or Cloud Code, which are platforms for automating tasks), as new ones will always emerge; instead, focus on learning valuable skills like problem-solving and clear communication.
Being "AI native" means automatically considering if AI can help with tasks, even if it only handles part of the job, as this mindset will save you time and keep you ahead of those who don't use AI at all.
The real value isn't in the AI itself (which everyone has equal access to), but in the unique systems, expertise, and experience you bring to using it, such as knowing what mistakes to avoid.

2026-07-31

Buzz is a free, open-source tool (like Slack) that lets you create teams of AI agents (like Codex and Claude Code) to work together on tasks.
It acts as a central place for all your context and information, allowing easy switching between different AI models and tools without losing history.
Buzz can connect to your existing AI agent accounts and use them to collaborate, like having Codex create a webpage and consulting with Claude Code.
It uses a protocol to communicate with these AI tools, injecting the necessary context for them to work together effectively.

2026-07-25

AI experts focus on using the right "tokens" (small pieces of text that AI models understand) for each task to improve quality and reduce costs.
Different AI models, like Chat GPT or Claude, have different strengths and weaknesses, and some may require more tokens to complete the same task.
Using two different AI models to review each other's work can help find more bugs and improve the quality of the output.
The speed at which AI models generate tokens (output speed) also matters, as faster models can save you time.

2026-07-22

**Model tiers** (different AI versions with unique costs, speeds, and abilities) are categorized into four types: Haiku (cheapest/fastest), Sonnet (balanced), Opus (deep reasoning), and Fable (most capable), each suited for specific tasks.
**Prompt engineering** (crafting clear instructions for AI) follows a step-by-step order: be clear, add context, use examples, structure with XML tags, assign roles, and control output format.
**Evaluation-first approach** means starting with the cheapest model tier that meets your needs, only upgrading if it can't handle the task, ensuring cost-effectiveness.
**Routing** (directing tasks to the right AI model) is like choosing a delivery service; match the task complexity to the model's capability to maintain quality, speed, and cost efficiency.

2026-07-19

Moonshot AI released Kimi K3, an open-source AI model (software anyone can use and modify) that's already available in apps like Kimi and Kimi Code, with impressive results for a low monthly cost.
Harbor SEO.ai, a new SEO (search engine optimization, helping websites rank higher on Google) tool for €29/month, scans websites for issues, suggests fixes, and even generates content.
Kimi K3 quickly built a well-designed, functional Irish Golf Directory website from a simple prompt, showcasing its potential for web development.
The model's capabilities, like generating SEO-friendly content and creating visual placeholders, make it a strong contender in open-source AI.

2026-07-16

Fable 5 (a new AI tool) can now handle entire business tasks independently, like building apps, creating videos, and market research, without constant human oversight.
It can build, test, and fix its own code, as well as create promotional videos and personalized emails, all based on voice commands.
Fable 5 can be used in different ways, like in an IDE (a tool for writing and testing code, like Cursor) or a browser, but it will become more expensive after July 7th.
WhisperFlow (a voice-to-text tool) is used to input commands to Fable 5, making the process faster and more accurate than typing.

2026-07-13

**New AI models:** OpenAI released GPT 5.6, including three models (Luna, Terra, and Soul), with Soul being the most advanced, designed for complex tasks.
**AI teamwork:** Combining Claude Fable 5 (Anthropic's AI, acting as an engineering manager) with OpenAI's Soul (as the engineer) can build real, useful software, like a sales training tool called Kendo.
**Security:** This AI team found and fixed security issues automatically, which could have been harmful if the product went live.
**Setup:** You can run these AI models on your own machine using a terminal and some simple commands, but it can be pricey due to the advanced capabilities.

2026-07-10

Hermes Agent (a powerful AI tool that can act like a full-time employee) works best with the Opus model (a specific AI model that's very reliable but expensive), but ChatGPT (a popular AI chat service) and GLM 5.2 (a cheaper AI model) are also options.
To avoid downtime, run at least two Hermes agents simultaneously, using different AI models or accounts, so they can monitor and fix each other if one fails.
You can create new Hermes agents (called "profiles") either by asking an existing agent to set one up for you or by using the Hermes dashboard.
If you're running a serious business, consider investing in the Opus model for Hermes Agent, as it's the most reliable for completing tasks.

Key points

What it is

**Prompt engineering** is crafting input text (called a prompt) to get useful output from an AI model, like giving it a role and clear instructions.
**System prompts** set rules and tone for the AI, while **context** is the specific data and expertise you feed it for better results.
**Tokens** are how AI reads text, like Scrabble tiles, with each word broken down into parts the AI can process.
**Hallucinating** is when AI makes up false information, and a safety net can help prevent this.

How to use it

Start with a **priming prompt** to outline the behavior you want, then give a clear task with context.
Be specific and push for actionable depth, asking the AI to give you something you can act on.
Use shorter, highly underspecified prompts to avoid overloading advanced models like Claude Fable 5.
Save tokens by using **system prompts**, which are set-once instructions that trigger every time.

Watch out for

Over-prompting with long, multi-step instructions can backfire with advanced models.
Avoid **vendor lock-in** by not pouring everything into one platform, as the most powerful model can get paused, priced out, or replaced.
Be concise to prevent the AI from running for hours or taking actions you never asked for.
Self-verifying agents can improve output quality but are expensive.

Tools named

Claude Fable 5 (an advanced AI model), LLM (a type of AI model that helps write better prompts)

Lesson 1: What is Prompt Engineering for AI and why it matters

Prompt engineering is the practice of crafting input text (called a prompt) to get a useful output from an AI model. In early 2025, this skill was crucial. You had to give the AI a role, clear instructions, the desired end state, examples, and explicit "do not" rules. Today, the value of prompt engineering is dropping fast. Models are so good that you can often just paste in a rough idea and the AI will help you write the prompt for itself (a technique called meta prompting).

Why does this matter for AI development? Simple prompts are a manual, one-shot process with no memory. Every run is done by hand, results vary each time, and if something breaks, a human must fix it. That's why developers now focus on systems (reusable, automated workflows) rather than isolated prompts. A system prompt (like studying the night before an exam) gives the model rules and tone, but real value comes from context (the specific data and expertise you feed it — a cheat sheet during the exam). The best AI setups combine a system prompt with rich, unique context that cannot be copied easily. In short: prompting alone is a low-skill, low-value task. The cutting-edge skill is thinking in systems and feeding models your own proprietary context.

Sources

Lesson 2: How to use Prompt Engineering for AI: step-by-step

How to Use Prompt Engineering for AI Step by Step

Prompt engineering is just how you ask an AI for something. Start by being specific. Think of ordering food: "Give me food" gets mystery meat, but "grilled salmon, medium, lemon on the side" gets exactly what you want. With AI, specificity works the same way.

Understand tokens (how AI reads text like Scrabble tiles). The word "hamburger" becomes three tokens: "ham," "burr," "g." The more tokens a model can handle, the more text it can work with at once.

To build a good prompt, first use a priming prompt (instructions at the start of a conversation) to outline behavior you want the AI to follow. Then give a clear task with context. For example: "Create a LinkedIn post for me about the new Claude Fable 5 model. Talk about why local AI models will become bigger. Tailor this to non-techie readers."

Your prompt should push for actionable depth. The AI often gives answers that are too thin or too detailed. Ask it to give you something you can act on. Also add a safety net to prevent hallucinating (AI lying to you). This works best when you give the AI a document to assess, like a contract or proposal.

Newer models like Claude Fable 5 can do too much — running for hours or taking actions you never asked for. Be concise. Don't write overly long prompts. If needed, let the AI spin up a mini team of agents, each with a clear single job, to check each other's work and return just the finished answer.

To save tokens, ask Claude to reduce token usage directly. You can also use an LLM to help you write better prompts. The key is to describe what you want clearly, then let the AI handle the details.

Sources

Lesson 3: Best practices and pitfalls

A common pitfall in prompt engineering is over-prompting: pushing the AI with long, multi-step instructions. With advanced models like Claude Fable 5, this backfires — it does too much, running for hours or claiming work it never touched. Instead, use shorter, highly underspecified prompts. For example, taking a screenshot of an error and just saying "fix it" often works fine.

Tokens (units of text the AI processes) are your resource. You don't want to pay for tokens for no reason. A key best practice is using system prompts, which are set-once instructions that trigger every time. For instance, a system prompt can define a casual, layman's tone for all your marketing posts. This avoids repeating the same instructions in every message, saving tokens.

Self-verifying agents are a powerful technique to improve output quality. Instead of one prompt with a plan living in Claude's memory, the agent attacks a problem from independent angles, then other agents try to refute what was found. Weak findings get torn apart, and the survivors converge — adversarial verification that runs until answers agree. This is expensive but effective.

Finally, avoid vendor lock-in. The most powerful model can get paused, priced out, or replaced. Build skills that work across tools, and don't pour everything into one platform.

Sources