References: What's next

Source material

Source curriculum (structural mirror, cited as further study):
• Full Stack Deep Learning, "LLM Bootcamp" (Spring 2023):
    What's Next?
  Instructors: Charles Frye, Sergey Karayev, and Josh Tobin
  Course page: https://fullstackdeeplearning.com/llm-bootcamp/
  Lecture videos: publicly available on the Full Stack Deep Learning
    YouTube channel
  License: bootcamp materials are published free to view but no explicit
    license (Creative Commons or otherwise) is published; lecture videos
    are on YouTube under standard terms.
  Required attribution: "Based on the structure of the Full Stack Deep
    Learning LLM Bootcamp (Spring 2023), by Charles Frye, Sergey Karayev,
    and Josh Tobin (fullstackdeeplearning.com/llm-bootcamp). This is an
    independent structural mirror in original prose; it reproduces no
    course materials, and Full Stack Deep Learning does not endorse it."
This lesson mirrors the structure of the corresponding bootcamp session (the field's direction)
and is taught as a survey: lighter pedagogy, breadth over depth, pointing
at the three deeper Phase 3 lessons that follow (training your own,
agents, industry perspective).

Watch this next

• Full Stack Deep Learning, LLM Bootcamp: What’s Next? by Charles Frye, Sergey Karayev, and Josh Tobin. The session this lesson mirrors. Note: the recorded version is from Spring 2023; the field has moved since (and the rest of this track’s lessons cover several of the moves), so use the recording as the structural source and read it alongside more recent material.

Going deeper

A short, durable list. Each link is a specific next step, not a generic pile.

• The Anthropic models page and [the equivalent for whichever provider you use most]. The canonical reference for current context lengths, multimodal capabilities, and per-token pricing across model families. The “longer context” and “smaller specialized model” directions show up directly here.

• The Hugging Face open-model leaderboards. A running view of the open-weight model landscape, including the smaller specialized models named in this lesson. Useful for sanity-checking the “smaller can beat frontier on narrow tasks” claim against current numbers.

• The Vercel AI SDK examples gallery (or the equivalent for your stack). A running survey of what production builders are actually shipping; new patterns show up here before they appear in formal courses.

Adjacent topics

Where this connects inside the track and the wider curriculum.

• LLM foundations for production (lesson 2). The three productive limits (context, cost, latency) are the constant lens every direction in this lesson is read through.

• Augmented language models (lesson 4). The tool-use loop introduced there scales up into agents (the deep dive in lesson 10).

• LLMOps (lesson 7). The regression-test discipline is what makes adopting any new capability (longer context, multimodal, new model, reasoning) safe rather than silently regressive.

• Training your own LLM (lesson 9). The deep dive on the “fine-tune an open model” point on the build-vs-buy spectrum.

• Agents (lesson 10). The deep dive on the agentic direction surveyed here.

• Track 15, lesson 14 (Reasoning and alignment, RLVR). The build-side companion for the reasoning-models direction: from-scratch / lab POV.

• Track 14, lesson 12 (Reasoning models and the road ahead). The using-side companion for the same direction.