#215 - Runway games, Meta Superintelligence, ERNIE 4.5, Adaptive Tree Search

Our 215th episode with a summary and discussion of last week's big AI news!

Recorded on 07/04/2025

Hosted by Andrey Kurenkov (twitter.com/andrey_kurenkov) and Jeremie Harris (twitter.com/jeremiecharris).

Feel free to email us your questions and feedback at contact@lastweekinai.comand/or hello@gladstone.ai

Read out our text newsletter and comment on the podcast at lastweekin.ai.

In this episode:

• Cloudflare's new AI data scraper blocking feature, its potential implications, and technical challenges

• Meta's aggressive recruitment for its Super Intelligence Labs division is covered, highlighting key hires from OpenAI and other leaders in the field

• Anthropic loses significant talent to Cursor, with details on their new economic futures program focusing on AI's impact on the labor market

• Notable open-source AI model releases from Baidu and Tencent are also discussed, including their performance metrics and potential applications.

Timestamps + Links:

• (00:00:11) Intro / Banter

• (00:01:43) News Preview

Tools & Apps

• (00:02:55) Cloudflare Introduces Default Blocking of A.I. Data Scrapers (nytimes.com/2025/07/01/technology/cloudflare-a…)

• (00:05:44) Runway is going to let people generate video games with AI (theverge.com/ai-artificial-intelligence/694531…)

• (00:11:24) Google embraces AI in the classroom with new Gemini tools for educators, chatbots for students, and more (techcrunch.com/2025/06/30/google-embraces-ai-i…)

• (00:16:23) No one likes meetings. They’re sending their AI note takers instead. (washingtonpost.com/technology/2025/07/02/ai-no…)

• (00:18:08) Google launches Doppl, a new app that lets you visualize how an outfit might look on you (techcrunch.com/2025/06/26/google-launches-dopp…)

• (00:19:14) Google's Imagen 4 text-to-image model promises 'significantly improved' boring images (engadget.com/ai/googles-imagen-4-text-to-image…)

Applications & Business

• (00:22:18) Mark Zuckerberg announces his AI ‘superintelligence’ super-group (theverge.com/news/695355/mark-zuckerberg-meta-…)

• (00:29:35) Anthropic Revenue Hits $4 Billion Annual Pace as Competition With Cursor Intensifies (theinformation.com/articles/anthropic-revenue-…)

• (00:35:10) As job losses loom, Anthropic launches program to track AI’s economic fallout (techcrunch.com/2025/06/27/as-job-losses-loom-a…)

• (00:38:04) OpenAI says it has no plan to use Google's in-house chip (finance.yahoo.com/news/openai-says-no-plan-goo…)

• (00:41:08) Nvidia stakes new startup that flips script on data center power (axios.com/2025/07/01/nvidia-startup-data-cente…)

• (00:44:11) TSMC Arizona Chips Are Reportedly Being Flown Back to Taiwan For Packaging; U.S. Semiconductor Supply Chain Still Remains Dependent on Taiwan (wccftech.com/tsmc-arizona-chips-are-reportedly…)

Projects & Open Source

• (00:46:57) Baidu releases open source model family ERNIE 4.5 (the-decoder.com/baidu-releases-open-source-mod…)

• (00:51:55) Tencent Open Sources Hunyuan-A13B: A 13B Active Parameter MoE Model with Dual-Mode Reasoning and 256K Context (marktechpost.com/2025/06/28/tencent-open-sourc…)

• (00:57:09) Together AI Releases DeepSWE: A Fully Open-Source RL-Trained Coding Agent Based on Qwen3-32B and Achieves 59% on SWEBench (marktechpost.com/2025/07/02/together-ai-releas…)

• (01:00:11) GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning (arxiv.org/abs/2507.01006)

• (01:04:10) DiffuCoder: Understanding and Improving Masked Diffusion Models for Code Generation (arxiv.org/abs/2506.20639v1)

Research & Advancements

• (01:06:21) Wider or Deeper? Scaling LLM Inference-Time Compute with Adaptive Branching Tree Search (arxiv.org/abs/2503.04412)

• (01:13:07) The Automated LLM Speedrunning Benchmark: Reproducing NanoGPT Improvements (arxiv.org/abs/2506.22419)

• (01:18:04) Claude 4 Opus and Sonnet reach 50%-time-horizon point estimates of about 80 and 65 minutes, respectively (metr.org/blog/2025-03-19-measuring-ai-ability-…)

• (01:21:37) Performance Prediction for Large Systems via Text-to-Text Regression (arxiv.org/abs/2506.21718)

• (01:25:38) Does Math Reasoning Improve General LLM Capabilities? Understanding Transferability of LLM Reasoning (arxiv.org/abs/2507.00432v1)

• (01:26:33) Correlated Errors in Large Language Models (arxiv.org/abs/2506.07962v1)

Policy & Safety

• (01:29:04) Forecasting Biosecurity Risks from LLMs (forecastingresearch.org/ai-enabled-biorisk)

• (01:36:06) AI Task Length Horizons in Offensive Cybersecurity (sean-peters-au.github.io/2025/07/02/ai-task-le…)

• (01:42:30) Inside Tech's Risky Gamble to Kill State AI Regulations for a Decade (obsolete.pub/p/inside-techs-risky-gamble-to-kill)

• (01:52:56) Denmark to tackle deepfakes by giving people copyright to their own features (theguardian.com/technology/2025/jun/27/deepfak…)

See Privacy Policy at art19.com/privacy and California Privacy Notice at art19.com/privacy#do-not-sell-my-info.