# Andrej Karpathy: 2025 LLM Year in Review

Source: https://www.youtube.com/watch?v=9HYyAktq_3Y
Recap page: https://rapidrecap.app/video/9HYyAktq_3Y
Generated: 2025-12-31T15:03:00.315+00:00

---
## Quick Overview

Andrej Karpathy's 2025 LLM Year in Review highlights that the major shift was not just in model size, but in the fundamental framework for building and interacting with LLMs, evidenced by the emergence of self-improving agents like RLRV (Reinforcement Learning from Verifiable Rewards) and the move towards local, specialized models over massive cloud-based ones.

**Key Points:**
- The most significant shift in LLM development in 2025 was a total realignment in the conceptual framework for building and interacting with these models, moving away from simple scaling.
- The paper by Carpathy introduces RLRV (Reinforcement Learning from Verifiable Rewards) as a crucial new ingredient that broke the previous formula of massive pre-training followed by supervised fine-tuning (SFT).
- RLVR models receive feedback tied to tasks (like solving math puzzles or coding challenges) that are verifiable, unlike subjective rewards used previously.
- The trend is moving towards local, specialized LLMs (like a 'banana demo' model) that run on user hardware, contrasting with the previous reliance on massive cloud-based models like those from OpenAI.
- The concept of 'agents' that can orchestrate complex LLM operations (like reading documents, summarizing, and generating reports) is becoming the standard deployment method.
- The shift means that programming is evolving from writing explicit code to developing sophisticated, specialized 'agents' that manage complex operational chains, often with a graphical user interface (GUI).

![Screenshot at 00:18: Karpathy introduces the key shift in LLM development, stating the headline is about the conceptual framework shift, not just models getting bigger.](https://ss.rapidrecap.app/screens/9HYyAktq_3Y/00-00-18.jpg)

**Context:** Andrej Karpathy reviews the state of Large Language Models (LLMs) as of 2025, focusing on paradigm shifts that go beyond mere model scaling. The discussion centers on new training methodologies, particularly Reinforcement Learning from Verifiable Rewards (RLRV), and the practical implications for how models are deployed and used, emphasizing a move toward local, specialized intelligence over generalized, cloud-hosted behemoths.

## Detailed Analysis

Andrej Karpathy summarizes that 2025 felt fundamentally transformative for Large Language Models (LLMs) because the industry shifted its conceptual framework for building and interacting with them, moving beyond the simple scaling of previous years. The major breakthrough discussed is Reinforcement Learning from Verifiable Rewards (RLRV), introduced by Carpathy, which replaces the older paradigm reliant on massive pre-training and supervised fine-tuning (SFT). RLRV models are trained using rewards derived from verifiable tasks, such as solving complex math problems or coding challenges, where the output can be objectively checked for correctness, unlike previous subjective reward signals. This is contrasted with models optimized only for subjective rewards or mimicking human thought processes. The practical result is that these new models, like the 'jaggies' (jagged) models, are becoming adept at reasoning and problem-solving without needing constant human feedback loops. Furthermore, there is a significant move away from massive, cloud-based models towards smaller, local models that run on consumer hardware, indicated by the success of models like 'banana demo' and the development of specialized GUIs for these local applications. The source suggests this shift has created a crisis of confidence in older benchmarks because the new models are demonstrating capabilities—like writing complex, multi-step code or orchestrating tasks across different domains (law, finance)—that were previously unattainable, effectively making the underlying code maintenance less critical than the orchestration layer.

### The 2025 Paradigm Shift

- Total realignment in the conceptual framework for building and interacting with LLMs
- Shift is away from simple scaling towards new methodologies like RLRV
- RLRV uses verifiable rewards for training.

### RLRV Mechanics

- Reinforcement Learning from Verifiable Rewards uses objective feedback (correctness of math/code) instead of subjective human feedback
- This allows models to develop strategies independently.

### Deployment Shift

- Moving away from massive, cloud-based LLMs to smaller, local models running on user hardware (e.g., 'banana demo')
- This shift is driven by the need for low latency and data privacy.

### The Agent Layer

- Models are now deployed as 'agents' that orchestrate complex tasks (e.g., reading reports, data parsing)
- These agents use custom tools and are managed by a GUI, making the underlying code less visible/necessary.

### New Benchmarks and Outcomes

- The new capabilities of these specialized models make old benchmarks (like those based on mimicking human thought) obsolete
- The source suggests these models are capable of complex reasoning tasks that previously required large teams of engineers.

![Screenshot at 00:09: Visual representation of the podcast setup, likely the host/narrator, signaling the start of the review.](https://ss.rapidrecap.app/screens/9HYyAktq_3Y/00-00-09.jpg)
![Screenshot at 00:33: Text overlay referencing the source paper that outlines the new paradigm shifts in LLM development.](https://ss.rapidrecap.app/screens/9HYyAktq_3Y/00-00-33.jpg)
![Screenshot at 01:21: Visual representation of the concept of Reinforcement Learning from Verifiable Rewards \(RLRV\) being discussed as a key ingredient.](https://ss.rapidrecap.app/screens/9HYyAktq_3Y/00-01-21.jpg)
![Screenshot at 02:04: Visual reference to the concept of 'gammable rewards signals' used in the RLRV framework.](https://ss.rapidrecap.app/screens/9HYyAktq_3Y/00-02-04.jpg)
![Screenshot at 04:16: Visual comparison illustrating that specialized, local LLMs are being optimized for survival/real-world tasks rather than just imitation.](https://ss.rapidrecap.app/screens/9HYyAktq_3Y/00-04-16.jpg)