Tony Holdstock-Brown on security as a job for a long-running agent: "what was good now might not be good in three months' time." Every company needs to be asking what its security loop is. Sponsored by Inngest: chainofthought.show/podcast/83-... 0 views 20/11/2026
Cloudera's AI query tool: a public model only vectorizes the question. Agents behind the wall run the SQL and Python. "AI is not allowed to see the data." It's important to decide what your agents have access to - and what they don't. youtu.be/wd_vf8btjmE 0 views 19/11/2026
Chain of Thought turned two last month and I almost missed it. 80+ episodes. Episode 1 was four people on a video call. Episode 83 had three 4K cameras and a couch. Thank you to every guest who has joined us, and to everyone listening. What should year three dig into? 0 views 18/11/2026
Tony Holdstock-Brown on personal agents: "Very cool, extremely scary." He shares the security worries, and still expects this to unfold very quickly over the next three to six months. Sponsored by Inngest: chainofthought.show/podcast/83-... 0 views 17/11/2026
Agree on what each result means before the agent runs the test. Andy Triedman's example: new classifier wins 95% of the time, ship. Under 50%, close the ticket. The gray area comes to a person. youtu.be/GabJ3JSpJeE 0 views 17/11/2026
Tony Holdstock-Brown's question for every product team: "how does it work if your user never sees your product overall?" Long-running personal agents are becoming how people use other products. Episode sponsored by Inngest: chainofthought.show/podcast/83-... 0 views 16/11/2026
Thank you to Walrus Memory, a season four presenting sponsor of Chain of Thought. Walrus Memory gives AI agents portable, verifiable memory that carries context across apps, sessions, and other agents. Get started: walrus.xyz/cot 0 views 16/11/2026
For Inngest, betting on agents has meant 93% activation for people who build with agents, up from ~40% 🤯 Most Inngest customers don't log into the UI anymore Their agents check in through the API, the CLI and MCP Tony Holdstock-Brown on Chain of Thought (sponsored by Inngest): youtu.be/1ioBprhHdSw 0 views 15/11/2026
"My processor got overloaded seeing the future of data analytics being automated." Cloudera's Shayde Christian on why that moment calls for a leadership commitment. youtu.be/wd_vf8btjmE 0 views 15/11/2026
Did the review pass? Did the user merge? Tony Holdstock-Brown's case for scoring agents from product events: "100% of your agents can be monitored." It covers every run, without a second LLM guessing Thanks to Inngest for sponsoring this conversation: chainofthought.show/podcast/83-... 0 views 14/11/2026
A skill validator should fail a broken skill, too. agent-skill-builder tests valid and invalid cases across Claude Code and portable Agent Skills profiles. It checks structure; you'll still need to evaluate the skill's output. github.com/conorbronsd... 0 views 13/11/2026
Claire Vo said Product Management Is Dead at Lenny's con When I talked to her in 2025 she expanded: the go-between PM, relaying between engineers, designers and executives, isn't a robust enough job - it's going away. Claire has been proven rather correct imo chainofthought.show/podcast/36-... 0 views 11/11/2026
Agent memory as a hierarchy: session traces, weekly priorities, canonical docs. Andy Triedman of Theory Ventures: chainofthought.show/podcast/79-... 0 views 10/11/2026
Static benchmarks can be memorized. You.com CTO Saahil Jain on why his team builds dynamic ones: chainofthought.show/podcast/81-... 0 views 10/11/2026
Competitors put dozens or hundreds of data engineers on "ChatGPT over the enterprise repository." Cloudera's Shayde Christian assigned seven, and asked a better question first: is this really an AI problem? youtu.be/wd_vf8btjmE 0 views 09/11/2026
Delighted to welcome CoralBricks as a Chain of Thought podcast sponsor. CoralBricks serves open models through one API, lightning fast and cheap inference. Get $10 in credits to try DeepSeek on your tasks: chainofthought.show/coralbricks 0 views 09/11/2026
What can AI do for you? Vercel CTO Malte Ubl encourage you to vibe check: paste your text and your rules into a chatbot and see if the model is anywhere up to the task. chainofthought.show/podcast/39-... 0 views 08/11/2026
You.com CTO Saahil Jain: model labs can train with their own harness in mind, so "their harness and the model become basically co-evolved together." The risk is overfitting. chainofthought.show/podcast/81-... 0 views 08/11/2026
New model generation: wipe your skills files or refactor them? DJ Sampath (Cisco) says wipe. I say refactor. DJ's example: "here's my personal PR bot. I'm gonna call it Jarvis. I think that's our problem right there." 0 views 08/11/2026
Delighted to welcome WisdomAI as a Chain of Thought podcast sponsor on our recent episodes. In their latest benchmark, Claude Code alone topped out at 73% accuracy. WisdomAI closed the gap, with accurate answers 3x faster at about a third of the cost. Read the report: wisdom.ai/benchmark 0 views 06/11/2026
"An agent by itself, it's just random numbers outputting text." Dan Lorenc of Chainguard on where agent risk actually comes from: the tools you bolt on. chainofthought.show/podcast/68-... 0 views 06/11/2026
Agents can run a thousand searches at once. Most people never make it to page two. Saahil Jain, CTO of You.com, on why search has to be rebuilt for AI agents. New on Chain of Thought: youtu.be/0Imi-BZVrsA 0 views 05/11/2026
A fleet of agents needs a shared way to stop work. Tyler Akidau on the governance layer to put in place before scaling. 0 views 04/11/2026
CVE to exploit in the wild: it used to be months. DJ Sampath, SVP at Cisco says it's now "sub-eight minutes - and it's only decreasing." "Your defenders need to start moving at agent speed." Learn how Cisco is adapting to the threat environment & securing personal agents for 90,000 employees 👇 0 views 31/10/2026
Cisco rolled out a personal AI agent to 90,000+ employees, with conditions: every prompt scanned, sandboxed agents, every skill people bring in checked SVP DJ Sampath on why security from the start sped up adoption - and using Codex to accelerate Full episode: chainofthought.show/podcast/80-... 0 views 30/10/2026
A person ignores an out-of-date PRD. An agent takes everything it reads as true. Theory Ventures' Andy Triedman on why stale docs become an agent problem, and why companies need a living source of truth. youtu.be/GabJ3JSpJeE 0 views 30/10/2026
A one-paragraph spec gets 3-4 follow-up questions from a coding agent. Andy Triedman says stopping there is one level too high: the next 8-12 questions decide the customer experience. youtu.be/GabJ3JSpJeE 0 views 29/10/2026
Agents are going to use your product 10x more than your customers did two years ago. Tony Holdstock-Brown, co-founder and CEO of Inngest, on what that does to your infrastructure. Full conversation soon on Chain of Thought. Subscribe: chainofthought.show/subscribe/ 0 views 28/10/2026
A software team that ships twice as much leaves competitors behind. A legal team doing twice the work doesn't. Theory Ventures' Andy Triedman on why agents get adopted fastest where the work is the product. youtu.be/GabJ3JSpJeE 0 views 28/10/2026
"The models are frankly all good enough." Theory Ventures' Andy Triedman: most agent work is context limited, not intelligence limited, and that's the opening for startups with deep domain expertise. youtu.be/GabJ3JSpJeE 0 views 27/10/2026
Same query, ClickHouse SQL vs Snowflake SQL: LLMs trip in different places. chainofthought.show/podcast/75-... Sponsored by WisdomAI. 0 views 26/10/2026
Kong never had webhook verification, because there were too many different webhook senders to support. Standard Webhooks changed that for Kong and ngrok. Tom Hacohen says Svix got a selfish win too: a standard to point to on sales calls. chainofthought.show/podcast/78-... 0 views 22/10/2026
Long-running AI agents are taking over - turning LLMs from stateless geniuses, to stateful problem solvers. Today, I joined Inngest Founder and CEO Tony Holdstock-Brown in SF to break it down for the Chain of Thought Podcast👇 Get our conversation in your inbox: newsletter.chainofthought.show/ 0 views 22/10/2026
2.5 months ago, I bet on myself & began building Chain of Thought full time. It's been a hectic year: two companies I helped build were acquired. Galileo by Cisco in April, Modular by Qualcomm in July My time at Modular ended shortly after acquisition - and I thought: what did I want to do next? 0 views 22/10/2026
Agents are becoming the primary users of many apps. Tom Hacohen: "Events are kinda unique in the sense that they're controlled by the sender, not the receiver." They arrive outside the agent's session, and laptops aren't always listening. chainofthought.show/podcast/78-... 0 views 20/10/2026
Sarah wrote it. I published it. But do I trust an AI agent to write and publish for me? chainofthought.show/podcast/75-... 0 views 19/10/2026
Local agents often get webhooks through a tunnel like ngrok. Close the laptop and the tunnel goes down. Tom Hacohen on where the events go: providers retry, give up, then hit you with a storm of retries when the lid opens. Svix's answer is polling. chainofthought.show/podcast/78-... 0 views 19/10/2026
Tom Hacohen calls JSON "an abomination." It's everywhere anyway, because it's easy to use. He puts webhooks in the same bucket: duct tape that's good enough and works. His bigger frustration is that protocols can't be scrapped and rebuilt. chainofthought.show/podcast/78-... 0 views 19/10/2026
When Svix launched on Hacker News, half the replies told Tom Hacohen he was wasting his life on a silly idea. The other half asked where he'd been a year ago, when they had to build webhooks themselves. He says "you need to have the haters." chainofthought.show/podcast/78-... 0 views 18/10/2026
People are 3D printing claws to keep laptops open so their agents keep running. When the lid closes anyway, Tom Hacohen says you come back to "a storm of retries from 12 hours ago." Svix lets the agent poll from the last event it got instead. chainofthought.show/podcast/78-... 0 views 18/10/2026
Webhooks run under a lot of the agentic internet, and agents are about to put far more load on them. Svix founder and CEO Tom Hacohen on what breaks when agents become the customer, from retry storms to whether agents still need SDKs. chainofthought.show/podcast/78-... 0 views 17/10/2026
Vercel CTO Malte Ubl, Sept 2025, on agents: "they all work." Now half of Vercel's deploys come from coding agents. chainofthought.show/podcast/39-... 0 views 15/10/2026
"That's not a 10% differential anymore, that's a 10x differential." Cisco's Jeetu Patel on the gap between people who are fluent with AI and people who aren't. chainofthought.show/podcast/67-... 0 views 14/10/2026
"Pets.com was like the poster child for the ridiculousness of 1999. Chewy.com is Pets.com. Chewy.com does great business." Laurie Voss: the dot-com optimists were wrong about timing, not direction. chainofthought.show/podcast/74-... 0 views 13/10/2026
Context OS: a Git-backed context and workflow layer for coding agents. Identity, projects, decisions and handoffs in plain Markdown, shared across Claude Code, Codex, OpenClaw and OpenCode. Try it: github.com/conorbronsd... 0 views 13/10/2026
"Many PRs, low cost, good. Less PRs, high cost bad." Sonar CTO Andrea Malagodi: those extremes hid two engineers solving different problems. chainofthought.show/podcast/76-... 0 views 13/10/2026
Chain of Thought is a technical AI podcast for engineers building with AI: conversations with the people building AI agents, infrastructure and developer tools. Plus, we explain AI topics for everybody. Check it out: chainofthought.show/?utm_source... 0 views 13/10/2026
0:37 "The people that removed lines of code were probably the most important people you had at the time." Sonar CTO Andrea Malagodi: token leaderboards echo the old pressure to produce more lines. Goodhart's law is prescient as always - full conversation: chainofthought.show/podcast/76-... 0 views 12/10/2026