Coding with AI

AI Systems Development

Last updated 2026-08-01

What's new

2026-08-01

Anthropic (the company behind the AI model) released a guide for Opus 5, their newest AI model, which tells you to simplify your prompts (the instructions you give the AI) by removing certain lines.
The guide suggests giving Opus 5 the entire task at once, rather than breaking it into steps, as this newer model works better with complete instructions.
It's important to clearly state what the AI should not do, to avoid it adding unnecessary work or content, which could waste your time and usage limits (the amount of data you can use with the AI).
When the AI finishes a task, it will tell you about it, but you should set limits on how much it can write in its reply and in the actual task it's completing.

2026-07-28

OpenAI's new ChatGPT voice feature lets you control AI agents (AI programs that can do tasks for you) with your voice, making it easier to use AI anywhere, anytime.
Unlike old voice features that just type what you say (called dictation), this new tool lets you command multiple AI agents across all your devices (like your phone, tablet, or computer) to get things done instantly.
ChatGPT voice can check the status of your projects, give you updates, and even create new AI agents to handle tasks for you, like fixing errors or building websites.
This tool can greatly increase your productivity (getting more done in less time) by removing barriers between having an idea and seeing it become reality.

2026-07-25

You can now use a smart AI tool (Local AI) that runs completely free, offline, and privately on your own computer, with no data leaving your device.
Local AI uses "open weights" (free, downloadable AI models) that you can own and use without paying for access or worrying about data privacy.
The free AI models have improved significantly, with some like GLM 5.2 (a large, capable AI model) performing close to top paid models on many tasks.
Using Local AI gives you control and privacy, as you're not renting a service from a company that can change or restrict access, and your data stays on your machine.

2026-07-19

AI is changing software, but it's risky because its outputs can be unpredictable, and users need to learn new ways to interact with it.
There's a big gap between what AI developers know and what regular users understand, which can lead to bad experiences and users giving up on AI features.
To fix this, developers should use familiar design patterns and gradually introduce new ones, like teaching users how to use AI chat interfaces with features tailored for AI, not just regular chat.

2026-07-13

Claude Code (a tool for building AI-powered automations) lets you work with local files and online services like Gmail, Slack, or a CRM (customer relationship management system), making it more powerful than Claude Chat (a simple AI chatbot).
Claude Code uses the same AI models (like Opus, Sonnet, or Haiku) as Claude Chat, but adds extra features for working with files and online services.
Claude Code is like an AI harness (a tool that helps you use AI models), which sits between the AI model (the engine) and you (the driver), helping you build automations and agents (AI systems that can do tasks for you).
The instructor, Nate, uses Claude Code to build and manage multiple businesses, showing how one person can do the work of a team with AI.

2026-07-01

Google DeepMind is already planning for Artificial Super Intelligence (ASI) (AI smarter than all humans combined), not just Artificial General Intelligence (AGI) (AI as smart as a typical human).
They predict that AGI could lead to ASI through scaling (bigger, better AI models) or algorithmic shifts (new AI architectures or training methods).
AI is advancing so fast that researchers are now writing papers with instructions for AI to summarize them, assuming AI will read them instead of humans.
The paper also discusses a theoretical "universal AI" (AIXI), the ultimate limit of AI intelligence, which we can approach but never truly reach.

2026-06-28

Paul Yushin and Luis François Bouchard teach how to build a personal AI research operating system (OS) to manage and retrieve notes from tools like Obsidian, Readwise, Notion, and Google Drive.
They emphasize creating a system that integrates with personal values and thoughts, using AI to pull relevant notes for current projects instead of manually searching.
The AI research OS is designed to be adaptable, allowing users to customize it for their specific needs, such as referencing past work to avoid repetition.
For complex problems, they recommend using a retrieval-augmented generation (RAG) pipeline with vector databases, but note that this requires infrastructure and is not as user-friendly for daily use.

2026-06-16

Google DeepMind, a leading AI company, released a 57-page paper titled "From AGI to ASI" (AGI (Artificial General Intelligence) is AI that performs like a typical human, while ASI (Artificial Superintelligence) is AI that outperforms all humans combined), outlining what happens after achieving human-level AI, with contributions from top AI researchers.
The paper defines four main pathways to reach ASI: pure scaling (bigger models, more data), algorithmic paradigm shifts (fundamentally different AI models), and recursive self-improvement (AI improving AI research).
The paper also introduces the concept of universal AI (AXI), a theoretical maximum intelligence that can't be practically achieved, similar to the speed of light in physics.
The authors wrote a section called "summary instructions" specifically for AI assistants, assuming they will summarize the paper for humans, indicating the growing role of AI in research.

2026-06-13

Anthropic (a company that makes AI tools) released a new AI model called Claude Mythos, also known as Fable 5, which is considered the most powerful AI model in the world for most tasks.
Fable 5 is exceptionally good at using tools, spatial reasoning, and creating visually oriented content, such as recreating a full presentation deck or designing a mobile app with just a few prompts.
The new model can build and run web and mobile apps, like a simple Minecraft game, by using other online services (like Daytona for creating a safe testing environment and Convex for managing data) with minimal input from the user.
The creator demonstrated building a functional notes app, similar to an existing app called Lovable, in just two prompts, showcasing the model's ability to quickly generate and improve upon complex designs.

2026-06-10

Learning one AI tool like Claude (a popular AI assistant) isn't wasted time because the skills you gain can transfer to other tools like Codex (a newer AI assistant).
AI tools like Claude, Codex, and Open Claw (different AI assistants) work similarly, using folders and context files on your computer, making it easy to switch between them.
Focus on understanding the fundamentals of AI tools, not just the specific tool, to avoid feeling overwhelmed by new releases and stay adaptable.
Your work in one AI tool can often be used in another, as they share similar structures and can access the same files and connected tools (like Gmail or Slack).

2026-06-07

**Multiple AI sessions**: Boris Sherny (the creator of Claude Code, an AI tool for coding) runs many AI sessions at once, each handling a single task, to boost productivity and avoid mixing contexts.
**Claude.md file**: This file stores rules and context for Claude Code, so you don't have to repeat instructions; it's like a cheat sheet that the AI checks every time it starts a session in that folder.
**Compound engineering loop**: By continuously updating the Claude.md file with new rules based on mistakes or lessons learned, the AI improves over time, making future sessions smarter and more efficient.
**Team collaboration**: Teams can share the same Claude.md file, so everyone benefits from the rules and improvements added by others, creating a shared knowledge base.

2026-06-04

Microsoft built seven new AI (artificial intelligence) models—like its own reasoning and coding brains—so it no longer relies only on partners' technology.
The new "MAI Thinking One" model cuts costs by up to 10 times, claims to match top rivals in quality, and uses legally clean training data.
"Microsoft IQ" is a new intelligence layer that plugs into company data and tools to make AI agents (AI programs that act on your behalf) less prone to mistakes and more helpful.

2026-06-03

A "hive mind" (shared memory system) lets multiple AI agents (autonomous helpers) work as a coordinated team and share what each learns.
Use /slash commands (text shortcuts like /standup) to coordinate all your agents at once through chat instead of managing each separately.
Connect AI agents to business tools through APIs (software bridges) and skills (new abilities), so they automatically inherit your entire digital infrastructure.
One agent can now manage real work like ad campaigns—analyzing performance data and automatically sending daily reports instead of manual tracking.

Key points

What it is

AI systems development is building software that learns from examples, not step-by-step instructions, like learning to cook by studying finished dishes instead of following a recipe.
AI has been around since 1956, but today it's more accessible and cheaper, making it important for businesses to use their unique data and processes to stand out.
Many automations don't need AI—about 50% of business automations are simple and can be done without it.
AI systems are only as smart as the data and expertise you feed them, and they should be built solidly and then improved over time.

How to use it

Start by clearly defining your project scope, including the problem you want to solve, the tools you'll use, and the expected outcome.
Combine multiple AI tools for a powerful workflow, using each tool for its strengths, like using Claude for research and Claude Code for building.
Use the agentic loop in Claude Code, which plans, acts, and observes results, and iterate within the same conversation if the first attempt isn't right.
Master one agentic coding tool, find high-value problems, and use it to create solutions, like using Claude Code to automate tasks through an API.

Watch out for

Treat AI as a mentor, not a vending machine, and always ask why before accepting AI output to prevent shipping broken code.
Understand that the harness (the setup and ecosystem around the model) matters more than the model itself, and give clear instructions and tools for AI to handle errors.
Avoid "dark code"—code you ship without understanding—and adopt the PIV loop to make AI coding predictable, breaking work into reusable, small chunks.

Tools named

Claude (an AI assistant from Anthropic), Claude Code (Anthropic's coding tool), Opus 4.7 (Anthropic's advanced model)

Lesson 1: What is AI Systems Development and why it matters

AI systems development is the process of building software that learns from examples instead of following step-by-step instructions. Think of it like cooking: traditional software follows a recipe exactly, but AI looks at thousands of finished dishes and writes its own recipe by figuring out the rules from the data you give it. The term "artificial intelligence" was coined in 1956, so the field is not new, but building these systems today matters for AI development because AI models are becoming cheaper and more accessible. Intelligence alone is becoming a commodity; what actually sets a successful AI system apart is your proprietary processes, decisions, and historical context. You have to collate that information and plug it into the right AI model with the right framework.

A key principle is that many automations can be simple and don't even need AI—about 50% of business automations fall into that category. When you do use AI, the system is only as smart as the data and subject matter expertise you feed it. A system prompt (instructions that set rules and tone) is like studying the night before an exam, but good context is like having a cheat sheet during the exam. The goal is to build something solid and then improve it as you learn how it behaves in production, because businesses change and AI models evolve. To make money with AI, you must stop selling AI agents and start selling AI solutions—diagnosing business problems and using AI to solve them, saving clients time, money, and focus.

Sources

Lesson 2: How to use AI Systems Development: step-by-step

To begin developing with AI systems, start by clearly defining your project scope. You must be able to explain what problem you want to solve, what tools you are using, and what the end result should look like. If you cannot communicate your plan clearly, neither a human nor an AI agent can build it for you.

A powerful workflow combines multiple AI tools rather than relying on just one. For example, use Claude (an AI assistant from Anthropic) to research a technology, then open Claude Code (Anthropic's coding tool) to build it. Claude gives you clarity; Claude Code gives you execution. After building, you can use another tool to generate release notes and stakeholder presentations. This full pipeline uses each tool for its strength.

When using Claude Code, understand the agentic loop (the cycle where the AI plans, acts, and observes results). It has built-in tools and safety nets. If the first attempt is not right, simply iterate within the same conversation—you do not need to start over. For larger codebases, Anthropic's playbook emphasizes that the harness (the setup and extension points around the model) matters more than the model itself.

You can also combine Claude Code with a scheduler called routines and a specific model like Opus 4.7 (Anthropic's advanced model) to automate tasks such as research, decision-making, and logging through an API. The key is mastering one agentic coding tool, finding high-value problems, and using Claude Code to create solutions.

Sources

Lesson 3: Best practices and pitfalls

When building AI systems with tools like Claude Code, beginners often fall into common traps. First, treat AI as a mentor, not a vending machine. Never accept AI output without asking why—this "curiosity rule" prevents you from shipping broken code. A study by Anthropic (the company behind Claude) found that developers using AI scored 17% lower on coding tests than those coding by hand. The same study showed experienced developers using AI took 19% longer to finish tasks, yet thought they were 24% faster. This gap between perception and reality is the real finding.

Second, understand that the harness (the ecosystem around the model) matters more than the model itself. Anthropic's Applied AI team published a playbook stating that how you set up your environment, instructions, and workflows decides how Claude Code performs. If something breaks, Claude Code will handle the error, research it, and adapt—but only if you've given it clear instructions and tools.

Third, avoid "dark code"—code you ship without understanding. Most developers treat AI coding like a slot machine. Instead, adopt the PIV loop (a system to make AI coding predictable). Break your work into reusable, small chunks. Remember that 90% of Claude Code was written by Claude Code itself, but that doesn't mean you should skip verification. Master one agentic coding tool (a tool that acts on its own to complete tasks) deeply, find the highest-leverage actions, and create solutions using that tool. A little education goes a long way.

Sources