Congratulations on getting your first job as a backend engineer, Opeyemi! 👏👏 This is a huge milestone and we're rooting for you! 🚀 Top dev stories on NewDev today. 🔥 #newdev

BREAKING: The US government just took down Claude Fable 5 and it's worse than it looks Full video: youtu.be/Ll0c6xZttdQ

Last week Anthropic begged the AI industry to slow down. A shared "brake pedal," they called it. This week they shipped the most powerful model they've ever released to the public. Claude Fable 5 is incredible. It's also wrapped in safeguards that say a lot about where this is all heading. I unpacked the whole story in my new video 👉 youtu.be/qipJ7umHjAs

Anthropic just released Claude Fable 5, its first generally available Mythos-class model, with Opus 4.8 fallback safeguards, 1M context, 128k output, and $10/$50 pricing. Claude Mythos 5 is the same underlying model as Fable 5, but with some safeguards removed. Fable 5 is the general-use product. Mythos 5 is the controlled-access version for organizations Anthropic believes need stronger cybersecurity or, later, biology capabilities under a trusted access program. Read more: getaibook.com/news/claude-fa…

They said I should stop making educational videos cause no one watches those anymore. Well, for the 2 of you who still do, in my latest video, we dive into LLMs and the technical details behind how they work. You'll learn: - The model behind ChatGPT - How text becomes tokens - Next-token prediction - What the model learns during training - Transformers and attention - Context windows - How responses are generated - AI Hallucination - Tools - How a model becomes an assistant Watch here 👉 youtu.be/MiuguTvf2Xk

It really sucks to be a software engineer right now. Heck, it sucks to be in almost any role at a tech company right now. The layoffs are exhausting. The job market feels strange. AI is changing the work faster than most people have had time to process. And a lot of people are wondering what happens next. I made a video about why this moment feels so hard, what's changing, and how to prepare for what's coming. Please watch here: youtu.be/sDh06gyRx2c Would love to hear how you're thinking about all of this.

Why does the same prompt give different answers on different days? Because the model isn't retrieving one fixed answer. It writes one token at a time. At each step, it has a set of likely next tokens, not a single guaranteed one. If the prompt is "write a polite reply...", the next few tokens might push it toward a short note, a warm apology, a formal email, a few bullet points, or a different example. More than one path can be reasonable. Once it chooses one path, that choice becomes part of the context for the next token. So a small difference near the start can compound into a noticeably different answer. That's why the same prompt can produce a different structure, example, variable name, level of detail, or tone. The answer isn't sitting there waiting to be fetched. Every run is built again, one token at a time.

Does ChatGPT actually "understand" what I'm saying? No, not the way a person does. A language model predicts the next token based on patterns it learned during training and the context it gets at runtime. That prediction can look like understanding because language has a lot of structure. If the model has seen enough examples of code, errors, docs, arguments, emails, and explanations, it can often continue the pattern in a useful way. But there's no private inner meaning behind it. It isn't reading your intent. It isn't forming beliefs. It isn't checking reality unless a tool or source gives it that information. That's why it can sound thoughtful one moment and invent something the next. It isn't understanding. It's very advanced pattern continuation.

Continued Pretraining: How specialiazed AI models are made

This is the easiest it’s ever been to look like a builder without building anything.

A prompt is a mirror of how well you understand the problem. Vague thinking in, vague output out. No framework fixes that. No template does either. You fix bad prompts by thinking harder before you type, not by adding more ritual to the prompt itself.

Your GitHub has: - A todo app from 2021 you built to "learn React." 1 commit. README says "more features coming soon." - A blog you started in 2022. 4 posts. The last one is titled "Why I'm Starting a Blog." You weren't. - A CLI tool that does something you've forgotten. The README is half-written in a tone you don't recognize as your own. - An open source project with 2 stars. Both are from your alt accounts. - A fork of a popular repo you were going to "contribute to." You never opened a PR. - A repo called "definitely-finishing-this." You didn't. Every January you say "this year I'm going to actually ship something." Then you start 4 new repos, mass-commit for a week, get distracted by a new idea, and the cycle repeats. The thing standing between you and the project that actually works is not talent. It's finishing.

Software is about to have a taste crisis.

I have a HUGE announcement! 🎉 My new book, Get Insanely Good at AI is live! 266 pages covering everything you need to actually be unstoppable with this technology. I've spent the last thirteen years building software, the last few using AI to build and ship things I could never have built alone, while teaching thousands of developers how to do the same. Everything I know about getting SERIOUS leverage out of this technology is in this book. It's live right now at getaibook.com. Please check it out. I'd love to hear back from you!

The industry moves forward one offended expert at a time.

The problem is not that AI writes code. It's that many people were only ever trained to contribute at the layer AI is now replacing.

The developer interview is stuck in a different era.

Google Launches Veo 3.1 Lite for Cheap AI Video Generation getaibook.com/news/google-la…

Anthropic accidentally leaked Claude Code's source map. Proprietary implementation details, context management, tool orchestration, and unreleased features, now public knowledge. Full architecture breakdown below 👇 getaibook.com/news/claude-co…

I strive to be like this dev everyday.

The moment you stop feeling like an imposter is the moment you stop growing.