iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
If you want to move beyond “type prompt, get answer,” start with an intuitive explanation of language models, then learn to build with them, and finally take on deeper academic material. That is the learning order Prayush Adhikari recommends in his September 27, 2026, DEV Community essay, based on three resources he watched. It is a useful map—not proof that this sequence is the best curriculum for everyone.
Three resources, three different learning goals
Adhikari’s guide groups the resources by what a learner hopes to do: understand the basic ideas, make software with language models, or study the subject more rigorously. The descriptions below reflect his account of the videos and course; they are not independent evaluations of their full contents.
| Resource | Best fit | Format and focus, as described by Adhikari |
|---|---|---|
| Piyush Garg, “How LLMs Works? – Overview” | Readers seeking an intuitive explanation of how LLMs work | Video overview of tokenization, embeddings, positional information, attention, next-token prediction, temperature, and training versus inference. Adhikari notes examples involving Hugging Face Transformers, Gemma tokenization, OpenAI embeddings, and PyTorch. |
| Dev Weekends, “Level 0 and Level 1 – Gen AI Sessions” | Readers who want to build applications using language models | Practical sessions described as covering API message roles, streaming, structured outputs, tool calling, prompting, and fine-tuning versus retrieval-augmented generation (RAG). |
| Stanford Online, “CME295: Transformers and Large Language Models I, Autumn 2025” | Readers looking for a more rigorous course of study | Adhikari describes a nine-lecture sequence addressing transformers, model families, training, tuning, reasoning, agents, evaluation, and current trends. The lecture count and syllabus are his account, not independently verified here. |
Start with the mental model: tokens, context, and prediction
For the first step, Adhikari recommends Garg’s overview as a way to get beyond surface-level familiarity and ask questions such as “what a token actually is.” In his account, the video connects tokenization and vector embeddings with positional information and attention, then explains generation as predicting subsequent tokens. It also distinguishes training, when model weights are updated, from inference, when a trained model produces output.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallThis gives newcomers a vocabulary for understanding model behavior before they encounter implementation details. It is an entry point rather than a complete technical treatment; the listed examples and topics are Adhikari’s description of the video.
#1 Best Overall
Move to application: learn the pieces of an LLM-powered feature
Once the basic concepts make sense, Adhikari points builders to the Dev Weekends sessions. His summary emphasizes the practical interface between an application and a model: message roles, streaming responses, structured outputs, and tool calling. It also includes prompting approaches and the choice between fine-tuning and RAG.
Where RAG fits
In the workflow Adhikari outlines, a RAG system parses documents, divides them into chunks, embeds and stores those chunks, retrieves relevant material for a query, and adds the retrieved text to the prompt. The point is to provide external context at answer time, rather than relying solely on what the model learned during training. He names RAGAS as an evaluation framework in this context.
RAG may reduce hallucination risk by grounding a response in retrieved material, but it does not eliminate that risk. Retrieval can miss relevant passages, and a model can still misread or misuse what it receives. These workshop details are the essay author’s description, not an independently verified session syllabus.
Then study the theory—and distinguish the Stanford courses
Adhikari’s third stop is Stanford Online’s CME295: Transformers and Large Language Models I, Autumn 2025. He characterizes it as a nine-lecture course moving across transformers, model families, training and tuning, reasoning, agents, evaluation, and current trends. Treat that outline and lecture count as his report of CME295, rather than as independently confirmed course details.
The Stanford material independently available in the sources for this guide is a different course: CS336: Language Models From Scratch (Spring 2025). Stanford describes CS336 as build-oriented, with lectures and assignments available online. It supports the broader point that Stanford offers public instruction in language models, but it does not verify CME295’s specific lecture count or syllabus. The CS336 course also cautions that lessons from small models do not necessarily transfer fully to frontier-scale models. Hands-on work can deepen understanding without being a full substitute for experience with models at that scale.
How to choose your starting point
- You want to understand the basic mechanics: Begin with Garg’s overview, as Adhikari recommends, and focus on the links between tokens, representations, attention, and next-token prediction.
- You want to make an application: Start with the Dev Weekends sessions and pay particular attention to API patterns, structured outputs, tool calling, and the fine-tuning/RAG distinction.
- You want a broad, deeper course: Look at the CME295 course Adhikari describes, while keeping its syllabus details attributed to his essay. Use Stanford’s CS336 materials as a separate, publicly available build-oriented learning option, not as confirmation of CME295.
- You are unsure: Follow Adhikari’s intuition-to-application-to-theory order as a reasonable path, then adjust based on whether your goal is conceptual understanding, software development, or more formal study.
What Adhikari’s weekend map is—and is not
The essay is a curated route through three learning resources, not a controlled comparison or a guarantee of learning outcomes. Adhikari writes: “If you’re on the same journey — curious about AI beyond "type prompt, get answer" — I’m breaking down what I learned from each one, and honestly, I’d watch them in the same order I did.” His sequence is best read as a practical recommendation grounded in his own viewing, with the Stanford course distinction kept clear.
Read Adhikari’s full DEV Community essay. See Stanford CS336: Language Models From Scratch (Spring 2025) lecture materials.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsQuick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
