26 posts tagged with "AI"

Agentic coding means you hand an AI agent a goal, not a list of steps. You say "make the failing user test pass" and the agent works out how: it reads the code, edits files, runs the commands, reads the errors, and tries again until the goal is met or it hits a limit. The word agentic marks the two things a plain chat window lacks, tools and its own decisions. A human still reviews the result. That last part is the part that matters for anyone learning to code, and October 2026 shipped real evidence for it.

MCP is a JSON-RPC 2.0 standard that lets an AI app (the host) call your code through a server exposing tools, resources and prompts, a typed API the model discovers at runtime. The 2026-07-28 revision made it stateless: no initialize handshake, no sessions. If you know APIs, the model and a STDIO server still hold; new code uses SDK v2.

A good Scrimba Explain prompt names the topic, the audience, and the slides you want. I ran the same RAG prompt through three teaching styles, attached a Markdown file, and built a three-lesson playlist, seven explainers in total. Professor was not noticeably deeper than Standard, which suggests the prompt text, not the picker, sets the depth.

If you do not know what Explain is yet, the reference page covers entry points and quotas, and the full review covers whether the output is any good. This post is only about the web composer at scrimba.com/explain↗ (opens in a new tab): what to type, which style to pick, and which clicks cost you an explainer.

Scrimba Explain exposes two MCP endpoints: one for your own agent session, one for pull requests. Claude Code connects to either with one command or one JSON file, reads your repository or diff, and streams a narrated video lesson into Scrimba. Drafts cost nothing until someone opens and claims the link. I ran both endpoints against a public Anthropic repo and a merged pull request, and this post is what came back, including the parts that went wrong.

Every Scrimba lesson has an EXPLAIN button in its header. Pause, type a question, and Scrimba builds a short narrated video from that lesson: its files, your highlighted code, and the teacher's narration to that point. It costs one explainer from your allowance, 10 for life on Free or 100 a month on Pro, and stays attached to the lesson.

Tested with a Pro account on the "Chunking text from documents" lesson in Learn RAG, Guil Hernandez's 94-minute course on retrieval-augmented generation, September 2026.

I pressed the button once, with a deliberately picky prompt, to see what it sends and what it builds. Here is what got sent, what came back, and the three moments in a course where it earns its cost.

Scrimba Explain turns a typed question into a narrated video with code, diagram and quiz slides in about ten seconds. NotebookLM Video Overviews turn documents already sitting in a notebook into a narrated slideshow that can take over 30 minutes. Pick Explain for a coding concept or a piece of code, NotebookLM for a video summary of your own PDFs or notes.

I generated every Explain result in this comparison myself, on a Pro account: the explainers, the transcripts, the quizzes. NotebookLM sat behind a corporate sign-in on the machine I had, so its side of the table comes from Google's own help pages and blog posts, linked where it appears, and anything Google does not document is marked unverified.

Scrimba Explain turns a question into a short narrated video explainer, complete with diagrams, code walkthroughs and, most of the time, a quiz at the end. A free Scrimba account gets 10 explainers for life, Pro gets 100 a month, and it works in the browser, inside Scrimba lessons, in ChatGPT, and inside coding agents like Claude Code through MCP. Most AI coding tools compete to write your code faster. This one explains it.

The best interactive coding platforms in 2026 are Scrimba, freeCodeCamp, Codecademy, Boot.dev, Exercism, and Codewars. Each one forces you to write and run real code instead of watching someone else do it, which is the whole game now that an AI can hand you a working answer on request. Pick by what stops you from copy-pasting, not by the brand name.

Scrimba and Codecademy suit guided frontend or fullstack learning best. freeCodeCamp wins on price. Boot.dev owns backend. Exercism and Codewars sharpen skills you already have. The sections below break down who each one actually fits.

The fastest way to get better at reading code is to read more of it on purpose: small pieces, out loud, predicting what each line does before you run it. AI writes the first draft now, so your job is verifying it, which makes reading the highest-leverage skill a self-taught developer can build in 2026. Most beginners still drill the half of the job the machine already does and skip the half interviewers now test.

I learned to code through 42 Network, a project-based school with no lectures, so this is written from the same seat as the readers who skip roadmaps and build.

You build coding experience without a job by manufacturing the proof a job would have given you: shipped projects real people use, open source contributions with your name on the merge, small freelance or volunteer work, and public build logs that show how you think. In 2026 that is not a clever workaround. It is the main road, because the entry-level rung that used to hand beginners their first experience has largely collapsed. I learned to code through the 42 Network rather than a CS degree or a bootcamp, so I built every hour of my early experience without an employer, and the tactics below are the ones that actually move a resume from ignored to interview.