<![CDATA[Weekly Claw]]>Sharp conversations with the people building AI: agents, devtools, infrastructure, startups, and the future of work. Hosted by Andy and Henry. New episodes Fridays. Explore every episode and send feedback at https://weeklyclaw.ai.

]]>https://weeklyclaw.aiRiverside.fm (https://riverside.com)Mon, 12 Oct 2026 21:08:16 GMTMon, 27 Jul 2026 19:36:47 GMT60Andy and Henry<p>Sharp conversations with the people building AI: agents, devtools, infrastructure, startups, and the future of work. Hosted by Andy and Henry. New episodes Fridays. Explore every episode and send feedback at <a rel="noopener noreferrer nofollow" href="https://weeklyclaw.ai" target="_blank">https://weeklyclaw.ai</a>.</p>episodicAndy and Henryhenrino3@gmail.comno<![CDATA[The Sandbox Failed | Weekly Claw #22]]>AI agents are escaping the boundaries we designed for them. In Weekly Claw #22, Andy and Henry unpack OpenAI’s Hugging Face security incident, the new Presence workspace, GPT-Live voice control, Cursor’s agent swarm rebuilding SQLite in Rust, Block’s open-source Buzz workspace, and Claude Opus 5.

They also debate whether AI labs or enterprises care more about cost and efficiency, why better context architecture can beat bigger swarms, Jensen Huang’s case for open models, and Sam Altman’s view of agentic engineering, token efficiency, and a future OpenAI IPO.

CHAPTERS

0:00 Welcome to Weekly Claw

0:50 The week AI agents broke assumptions

3:10 OpenAI’s agent escapes the sandbox

8:12 OpenAI Presence for enterprise agents

9:38 GPT-Live and voice-driven work

13:31 Cursor’s agent swarm rebuilds SQLite

16:20 Block Buzz: an open agent workspace

19:23 Claude Opus 5 and the price of intelligence

22:34 Jensen Huang, open models, and competition

28:23 Sam Altman on agentic engineering and IPOs

32:36 Hot take: cost, quality, and model efficiency

38:29 Heritage Telecom

38:50 What to watch next

40:16 Feedback and closing

41:10 Weekly Claw outro

SOURCES & CORRECTIONS

OpenAI — Hugging Face model evaluation security incident

https://openai.com/index/hugging-face-model-evaluation-security-incident/

OpenAI — Introducing Presence

https://openai.com/index/introducing-openai-presence/

OpenAI — ChatGPT Enterprise and Edu release notes

https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes

Cursor — Agent swarm model economics

https://cursor.com/blog/agent-swarm-model-economics

Block — Buzz

https://buzz.xyz

Anthropic — Claude Opus

https://www.anthropic.com/claude/opus

NVIDIA — The future of AI is open and proprietary

https://blogs.nvidia.com/blog/ai-future-open-and-proprietary/

CNBC — Sam Altman interview transcript

https://pressroom.versantmedia.com/cnbc/press-releases/cnbc-exclusive-transcript-openai-ceo-sam-altman-speaks-cnbcs-julia-boorstin

Correction at 19:57: Claude Opus 5 costs $5/$25 per million input/output tokens—50% below Fable 5’s $10/$50, not 20%.

Weekly Claw is a technology and AI podcast about agents, devtools, infrastructure, startups, and the future of software.

Watch every episode and send feedback:

https://weeklyclaw.ai

Supported by Heritage Telecom:

https://heritagetell.com

#AI #AIAgents #OpenAI #Cursor #Claude #WeeklyClaw

]]>bbf74cd6-167a-4a49-8f70-36ca093d22cbMon, 27 Jul 2026 19:41:17 GMT<p>AI agents are escaping the boundaries we designed for them. In Weekly Claw #22, Andy and Henry unpack OpenAI’s Hugging Face security incident, the new Presence workspace, GPT-Live voice control, Cursor’s agent swarm rebuilding SQLite in Rust, Block’s open-source Buzz workspace, and Claude Opus 5.</p><p>They also debate whether AI labs or enterprises care more about cost and efficiency, why better context architecture can beat bigger swarms, Jensen Huang’s case for open models, and Sam Altman’s view of agentic engineering, token efficiency, and a future OpenAI IPO.</p><p>CHAPTERS</p><p>0:00 Welcome to Weekly Claw</p><p>0:50 The week AI agents broke assumptions</p><p>3:10 OpenAI’s agent escapes the sandbox</p><p>8:12 OpenAI Presence for enterprise agents</p><p>9:38 GPT-Live and voice-driven work</p><p>13:31 Cursor’s agent swarm rebuilds SQLite</p><p>16:20 Block Buzz: an open agent workspace</p><p>19:23 Claude Opus 5 and the price of intelligence</p><p>22:34 Jensen Huang, open models, and competition</p><p>28:23 Sam Altman on agentic engineering and IPOs</p><p>32:36 Hot take: cost, quality, and model efficiency</p><p>38:29 Heritage Telecom</p><p>38:50 What to watch next</p><p>40:16 Feedback and closing</p><p>41:10 Weekly Claw outro</p><p>SOURCES &amp; CORRECTIONS</p><p>OpenAI — Hugging Face model evaluation security incident</p><p><a rel="noopener noreferrer nofollow" href="https://openai.com/index/hugging-face-model-evaluation-security-incident/" target="_blank">https://openai.com/index/hugging-face-model-evaluation-security-incident/</a></p><p>OpenAI — Introducing Presence</p><p><a rel="noopener noreferrer nofollow" href="https://openai.com/index/introducing-openai-presence/" target="_blank">https://openai.com/index/introducing-openai-presence/</a></p><p>OpenAI — ChatGPT Enterprise and Edu release notes</p><p><a rel="noopener noreferrer nofollow" href="https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes" target="_blank">https://help.openai.com/en/articles/10128477-chatgpt-enterprise-edu-release-notes</a></p><p>Cursor — Agent swarm model economics</p><p><a rel="noopener noreferrer nofollow" href="https://cursor.com/blog/agent-swarm-model-economics" target="_blank">https://cursor.com/blog/agent-swarm-model-economics</a></p><p>Block — Buzz</p><p><a rel="noopener noreferrer nofollow" href="https://buzz.xyz" target="_blank">https://buzz.xyz</a></p><p>Anthropic — Claude Opus</p><p><a rel="noopener noreferrer nofollow" href="https://www.anthropic.com/claude/opus" target="_blank">https://www.anthropic.com/claude/opus</a></p><p>NVIDIA — The future of AI is open and proprietary</p><p><a rel="noopener noreferrer nofollow" href="https://blogs.nvidia.com/blog/ai-future-open-and-proprietary/" target="_blank">https://blogs.nvidia.com/blog/ai-future-open-and-proprietary/</a></p><p>CNBC — Sam Altman interview transcript</p><p><a rel="noopener noreferrer nofollow" href="https://pressroom.versantmedia.com/cnbc/press-releases/cnbc-exclusive-transcript-openai-ceo-sam-altman-speaks-cnbcs-julia-boorstin" target="_blank">https://pressroom.versantmedia.com/cnbc/press-releases/cnbc-exclusive-transcript-openai-ceo-sam-altman-speaks-cnbcs-julia-boorstin</a></p><p>Correction at 19:57: Claude Opus 5 costs $5/$25 per million input/output tokens—50% below Fable 5’s $10/$50, not 20%.</p><p>Weekly Claw is a technology and AI podcast about agents, devtools, infrastructure, startups, and the future of software.</p><p>Watch every episode and send feedback:</p><p><a rel="noopener noreferrer nofollow" href="https://weeklyclaw.ai" target="_blank">https://weeklyclaw.ai</a></p><p>Supported by Heritage Telecom:</p><p><a rel="noopener noreferrer nofollow" href="https://heritagetell.com" target="_blank">https://heritagetell.com</a></p><p>#AI #AIAgents #OpenAI #Cursor #Claude #WeeklyClaw</p>no00:41:21122The Sandbox Failed | Weekly Claw #22full<![CDATA[The model is portable. The workflow is the moat.]]>AI software factories are no longer theoretical. In Weekly Claw #21, Henry and Andy are joined by Vorflux founder Haritha Elango to see how cloud coding agents plan, build, test, review and merge real software.

They unpack Kimi K3, GPT-RED and Thinking Machines Lab’s Inkling, then break down the AI Engineer World’s Fair, Peter Steinberger’s inner-loop/outer-loop model, the rise of harness engineering, and whether skills are becoming the next form of technical debt.

CHAPTERS

0:00 Welcome to Weekly Claw

1:35 OpenAI agent safety and destructive actions

3:20 GPT-RED: automated red teaming

3:53 Inkling and enterprise-owned AI

8:02 Kimi K3 and the open-weight race

10:46 Why agent loops matter

12:44 Vorflux: the AI software factory

19:28 How Vorflux began

22:12 Cloud environments and mobile testing

24:29 Building Vorflux with AI

27:31 Founder Q&A wrap-up

28:43 Recovering the live show

35:07 AI Engineer World’s Fair: software factories

37:12 Peter Steinberger and the outer loop

43:28 Hot take: software engineering becomes factory engineering

49:56 Skills hell and technical debt

57:54 MCP vs durable Linux CLIs

58:46 MC Porter and CLI-first tooling

59:25 Weekly Claw outro

Weekly Claw is a technology and AI podcast covering agentic software, open models, AI infrastructure and the people building at the frontier.

Follow Weekly Claw: https://weeklyclaw.ai

Sponsors:

Herald Labs — https://labs.theherald.co

Heritage Telecom — https://heritagetel.com

#AI #AIAgents #SoftwareEngineering #KimiK3 #Vorflux #WeeklyClaw

]]>ec739dec-f852-4d9d-86da-de2816ca7ba6Mon, 27 Jul 2026 19:40:10 GMT<p>AI software factories are no longer theoretical. In Weekly Claw #21, Henry and Andy are joined by Vorflux founder Haritha Elango to see how cloud coding agents plan, build, test, review and merge real software.</p><p>They unpack Kimi K3, GPT-RED and Thinking Machines Lab’s Inkling, then break down the AI Engineer World’s Fair, Peter Steinberger’s inner-loop/outer-loop model, the rise of harness engineering, and whether skills are becoming the next form of technical debt.</p><p>CHAPTERS</p><p>0:00 Welcome to Weekly Claw</p><p>1:35 OpenAI agent safety and destructive actions</p><p>3:20 GPT-RED: automated red teaming</p><p>3:53 Inkling and enterprise-owned AI</p><p>8:02 Kimi K3 and the open-weight race</p><p>10:46 Why agent loops matter</p><p>12:44 Vorflux: the AI software factory</p><p>19:28 How Vorflux began</p><p>22:12 Cloud environments and mobile testing</p><p>24:29 Building Vorflux with AI</p><p>27:31 Founder Q&amp;A wrap-up</p><p>28:43 Recovering the live show</p><p>35:07 AI Engineer World’s Fair: software factories</p><p>37:12 Peter Steinberger and the outer loop</p><p>43:28 Hot take: software engineering becomes factory engineering</p><p>49:56 Skills hell and technical debt</p><p>57:54 MCP vs durable Linux CLIs</p><p>58:46 MC Porter and CLI-first tooling</p><p>59:25 Weekly Claw outro</p><p>Weekly Claw is a technology and AI podcast covering agentic software, open models, AI infrastructure and the people building at the frontier.</p><p>Follow Weekly Claw: <a rel="noopener noreferrer nofollow" href="https://weeklyclaw.ai" target="_blank">https://weeklyclaw.ai</a></p><p>Sponsors:</p><p>Herald Labs — <a rel="noopener noreferrer nofollow" href="https://labs.theherald.co" target="_blank">https://labs.theherald.co</a></p><p>Heritage Telecom — <a rel="noopener noreferrer nofollow" href="https://heritagetel.com" target="_blank">https://heritagetel.com</a></p><p>#AI #AIAgents #SoftwareEngineering #KimiK3 #Vorflux #WeeklyClaw</p>no00:59:35121The model is portable. The workflow is the moat.full<![CDATA[The Cost of Intelligence Collapsed: GPT-5.6, Grok 4.5 & Agentic Ransomware | Weekly Claw #20]]>The cost of AI intelligence just collapsed. In Weekly Claw #20, Andy and Henry unpack the frontier-model pile-up around GPT-5.6, Grok 4.5, Claude Cowork, Cursor Automations and rapidly falling agent costs.

They debate whether models are becoming commodities, why enterprise routing is harder than it looks, whether companies should run DeepSeek in production, and what the first autonomous agentic ransomware attack means for every team building agents. The episode also covers the Novita × Kilo Code hackathon, Jensen Huang and Harrison Chase on open agent systems, and a hands-on GPT-5.6 vs Fable model comparison.

CHAPTERS

0:00 The model pile-up

2:42 Why intelligence costs collapsed

4:59 Frontier tiers and model pricing

8:19 Claude Cowork goes remote

9:23 Cursor Automations and always-on agents

10:44 Chinese models reshape enterprise AI

12:32 Three waves toward sovereign AI

16:05 Autonomous ransomware arrives

19:00 Novita × Kilo Code hackathon

25:07 Hot take: is the model moat gone?

31:21 Would you run DeepSeek in production?

37:52 Jensen Huang on open agent systems

40:43 Tool fight: GPT-5.6 vs Fable and Grok

49:21 What to watch next

50:13 Weekly Claw updates and feedback

Weekly Claw is a technology and AI podcast about agents, devtools, infrastructure and the future of work.

Explore every episode: https://weeklyclaw.ai

Follow Weekly Claw on X: https://x.com/weeklyclaw

Supported by:

Ada at Superada — https://superada.ai

Herald Labs — https://labs.theherald.co

HeritageTel — https://heritagetel.com

OpenClaw — https://openclaw.ai

#AI #AIAgents #GPT56 #Grok45 #AgentSecurity #WeeklyClaw

]]>0e15570b-a279-42a4-929e-a814e75cfecaMon, 27 Jul 2026 19:38:31 GMT<p>The cost of AI intelligence just collapsed. In Weekly Claw #20, Andy and Henry unpack the frontier-model pile-up around GPT-5.6, Grok 4.5, Claude Cowork, Cursor Automations and rapidly falling agent costs.</p><p>They debate whether models are becoming commodities, why enterprise routing is harder than it looks, whether companies should run DeepSeek in production, and what the first autonomous agentic ransomware attack means for every team building agents. The episode also covers the Novita × Kilo Code hackathon, Jensen Huang and Harrison Chase on open agent systems, and a hands-on GPT-5.6 vs Fable model comparison.</p><p>CHAPTERS</p><p>0:00 The model pile-up</p><p>2:42 Why intelligence costs collapsed</p><p>4:59 Frontier tiers and model pricing</p><p>8:19 Claude Cowork goes remote</p><p>9:23 Cursor Automations and always-on agents</p><p>10:44 Chinese models reshape enterprise AI</p><p>12:32 Three waves toward sovereign AI</p><p>16:05 Autonomous ransomware arrives</p><p>19:00 Novita × Kilo Code hackathon</p><p>25:07 Hot take: is the model moat gone?</p><p>31:21 Would you run DeepSeek in production?</p><p>37:52 Jensen Huang on open agent systems</p><p>40:43 Tool fight: GPT-5.6 vs Fable and Grok</p><p>49:21 What to watch next</p><p>50:13 Weekly Claw updates and feedback</p><p>Weekly Claw is a technology and AI podcast about agents, devtools, infrastructure and the future of work.</p><p>Explore every episode: <a rel="noopener noreferrer nofollow" href="https://weeklyclaw.ai" target="_blank">https://weeklyclaw.ai</a></p><p>Follow Weekly Claw on X: <a rel="noopener noreferrer nofollow" href="https://x.com/weeklyclaw" target="_blank">https://x.com/weeklyclaw</a></p><p>Supported by:</p><p>Ada at Superada — <a rel="noopener noreferrer nofollow" href="https://superada.ai" target="_blank">https://superada.ai</a></p><p>Herald Labs — <a rel="noopener noreferrer nofollow" href="https://labs.theherald.co" target="_blank">https://labs.theherald.co</a></p><p>HeritageTel — <a rel="noopener noreferrer nofollow" href="https://heritagetel.com" target="_blank">https://heritagetel.com</a></p><p>OpenClaw — <a rel="noopener noreferrer nofollow" href="https://openclaw.ai" target="_blank">https://openclaw.ai</a></p><p>#AI #AIAgents #GPT56 #Grok45 #AgentSecurity #WeeklyClaw</p>no00:53:23120The Cost of Intelligence Collapsed: GPT-5.6, Grok 4.5 & Agentic Ransomware | Weekly Claw #20full