Resources
The reading list — curated, opinionated.
30 resources worth your time: foundational papers, the few books that actually hold up, the blogs that catch ship-change moments, and the courses that pay off.
Attention Is All You Need
Vaswani et al. · 2017
The transformer paper. Read once, return when something doesn't make sense.
paperTransformersLanguage Models are Few-Shot Learners (GPT-3)
Brown et al. · 2020
The paper that made "scale is enough" the dominant frame in LLM research.
paperFoundationsTraining language models to follow instructions with human feedback (InstructGPT)
Ouyang et al. · 2022
How chat assistants stopped feeling like autocomplete and started being useful.
paperFine-tuningDirect Preference Optimization
Rafailov et al. · 2023
DPO: alignment without the PPO loop. The default for open-source alignment now.
paperFine-tuningRetrieval-Augmented Generation for Knowledge-Intensive NLP
Lewis et al. · 2020
The original RAG paper. Read it to understand why the pattern stuck.
paperRAGLost in the Middle
Liu et al. · 2023
Why dumping 20 chunks into the context doesn't work. Drives the case for reranking.
paperRAGReAct: Synergizing Reasoning and Acting in Language Models
Yao et al. · 2022
The pattern under almost every modern agent. Read once, internalize forever.
paperAgentsToolformer: Language Models Can Teach Themselves to Use Tools
Schick et al. · 2023
Why tool-use training shifted from a research curiosity to a default capability.
paperAgentsLoRA: Low-Rank Adaptation of Large Language Models
Hu et al. · 2021
The technique that made fine-tuning consumer-GPU practical.
paperFine-tuningQLoRA: Efficient Finetuning of Quantized LLMs
Dettmers et al. · 2023
LoRA + 4-bit quantization. Single H100 fine-tunes 70B models. Game changer.
paperFine-tuningDeep Learning
Goodfellow, Bengio, Courville · 2016
The textbook. Free online. Skim Part I if you're starting; reference Part II as needed.
bookFoundationsHands-On Large Language Models
Jay Alammar & Maarten Grootendorst · 2024
Best 2024 book on practical LLMs. Pairs well with reading the original papers.
bookFoundationsDesigning Machine Learning Systems
Chip Huyen · 2022
The MLOps book. Vendor-agnostic, principles-first, still relevant.
bookMLOpsAI Engineering
Chip Huyen · 2024
The successor focused on the post-foundation-model era. Best single intro to the AI engineer role.
bookMLOpsThe Illustrated Transformer
Jay Alammar
The post that made attention click for thousands of engineers. Still the best visualization.
blogTransformers3Blue1Brown — But what is a neural network?
3Blue1Brown
Visual intuition for what gradients, layers, and attention actually do. No math required to start.
talkFoundationsAnthropic's prompt engineering interactive tutorial
Anthropic
Best free prompt-engineering material from a frontier lab. Working code throughout.
docsPromptingHugging Face NLP Course
Hugging Face
Free, hands-on, code-first. The pragmatist's alternative to a degree.
courseFoundationsEugene Yan — engineering blog
Eugene Yan
Production ML / AI patterns from a practitioner. The "writing a system design doc" energy.
blogMLOpsLilian Weng — research blog
Lilian Weng
Long-form, deeply-cited surveys of attention, agents, prompting, alignment. Best for a "fill in my background knowledge" weekend.
blogFoundationsKarpathy — Neural Networks: Zero to Hero
Andrej Karpathy
Build a tokenizer, train GPT-2 from scratch, write the autograd engine. The masterclass.
courseFoundationsSimon Willison's Weblog
Simon Willison
Daily-ish thoughts on LLMs, prompt injection, tool use. Excellent radar for what just shipped.
blogAgentsHamel Husain — evals
Hamel Husain
The eval-discipline blog. If you read one thing on AI evaluation, read this.
blogMLOpsAnthropic's "Engineering at Anthropic" blog
Anthropic
Prompt caching, agent design patterns, and other guidance directly from the lab.
blogMLOpsOpenAI Cookbook
OpenAI
Ground-truth examples for the OpenAI API. Even if you use Anthropic, the patterns translate.
docsPromptingConcrete Problems in AI Safety
Amodei et al. · 2016
The "what could go wrong" taxonomy that still frames most safety conversations.
paperSafetyAnthropic's Responsible Scaling Policy
Anthropic
How a frontier lab thinks about deployment thresholds. Worth reading for the framing.
docsSafetyOWASP Top 10 for LLM Applications
OWASP
The application-security checklist for LLM features. Read before shipping any user-facing AI.
docsSafetyHow to switch into AI/ML — a no-BS guide
Vicki Boykis · 2022
Honest, opinionated transition guide. The opposite of "10x your career in 30 days" content.
blogCareerThe State of AI Report
Air Street Capital
Annual industry survey. Useful for "where is hiring", "what trends are real" calibration.
docsCareer