# #234 - Opus 4.6, GPT-5.3-codex, Seedance 2.0, GLM-5

Page: https://stenobird.com/podcast/last-week-in-ai/234-opus-4-6-gpt-5-3-codex-seedance-2-0-glm-5
Text version: https://stenobird.com/podcast/last-week-in-ai/234-opus-4-6-gpt-5-3-codex-seedance-2-0-glm-5.md
Podcast: [Last Week in AI](https://stenobird.com/podcast/last-week-in-ai)
Published: 2026-02-16T07:30:00+00:00
Episode link: https://rss.art19.com/episodes/a2a3c8c8-8c7d-4ddd-a277-4a0ee91f1e81.mp3?rss_browser=BAhJIg90cmFuc2NyaWJyBjoGRVQ%3D--952c5701c84ad333c69d5faa668f8177091704f0
Audio file: https://rss.art19.com/episodes/a2a3c8c8-8c7d-4ddd-a277-4a0ee91f1e81.mp3?rss_browser=BAhJIg90cmFuc2NyaWJyBjoGRVQ%3D--952c5701c84ad333c69d5faa668f8177091704f0
Processing state: processed
JSON: https://stenobird.com/v1/public/podcasts/last-week-in-ai/episodes/234-opus-4-6-gpt-5-3-codex-seedance-2-0-glm-5
Duration seconds: 5433

## Resource

A deep dive into the massive wave of new model releases from Anthropic, OpenAI, and Google, alongside the rise of high-performance Chinese models. The episode explores the shift from pure scaling to agentic workflows and the emerging risks of model incoherence.

## Highlights
- Main idea: Anthropic's Opus 4.6 marks a transition from simple chat to 'agent teams' with massive context windows
- Practical takeaway: OpenAI's GPT-5.3 Codex is leveraging Cerebras hardware to achieve significant speed increases for coding tasks
- Failure mode: Increasing model scale may lead to 'incoherence,' where models become more capable but more unpredictable and jittery
- Market trend: Massive venture capital inflows, such as ElevenLabs' $11B valuation, signal a rapid maturation of the AI startup lifecycle
- Technical shift: The industry is moving toward hybrid architectures, like Qwen3 Coder, to balance efficiency with agentic capabilities

## Topics

Anthropic Opus, OpenAI GPT-5.3, Google Gemini 3, Generative Video, AI Agents, Large Language Models, Cerebras, Model Incoherence

## Chapters
- 9:00 — Anthropic's Shift to Agentic Workflows: Discussion on how Opus 4.6 is being positioned as a universal knowledge worker rather than just a developer tool.
- 16:10 — OpenAI, Codex, and the Cerebras Advantage: Analyzing the performance gains of GPT-5.3 Codex and why the industry is looking for alternatives to the NVIDIA stack.
- 30:35 — Google Gemini 3 and STEM Benchmarks: Evaluating Gemini 3 Deep Think's performance in science and engineering despite concerns over safety documentation.
- 37:50 — The Rise of Chinese Generative Media: A look at ByteDance's Seedance 2.0 and Alibaba's Qwen Image 2.0 as serious competitors in the video and image space.
- 51:25 — The Economics of AI Scaling: Examining the massive valuations of ElevenLabs and Runway and the sustainability of high-margin AI services.
- 1:12:25 — The Problem of Model Incoherence: A technical discussion on how scaling compute might increase capability while simultaneously increasing task-level jitter and unpredictability.

## Actions

- request_transcript: `POST https://stenobird.com/v1/public/podcasts/last-week-in-ai/episodes/234-opus-4-6-gpt-5-3-codex-seedance-2-0-glm-5/transcription-requests` — Idempotently request low-priority transcript generation for this episode.
- read_markdown: `GET https://stenobird.com/podcast/last-week-in-ai/234-opus-4-6-gpt-5-3-codex-seedance-2-0-glm-5.md` — Read the agent-friendly Markdown representation of this episode resource.

A page view does not enqueue transcription. Agents should invoke `request_transcript` explicitly when they need this episode processed.

## Transcript

Full transcripts are not published on public pages unless there is a clear rights basis.