Nick Kuhn is scheduled to speak at the Midwest AI Summit on October 15 in Indianapolis about deploying AI agents in real-world environments.
“Nick is gonna come talk about what it looks like to deploy AI agents in real world environments.”
Person
Daniel Whitenack
24 episodes · 131 quoted claims
- 24 Sep 2026From AGENTS.md to Enterprise Deployment48m
- 17 Sep 2026How to get discovered in AI search55m
- 25 Aug 2026AI Proficiency: From Users to Builders56m
- 6 Aug 2026Models, Harnesses, and Multi-Agent Systems49m
- 23 Jul 2026Surviving the New Economics of a Post-Agentic World35m
- 9 Jul 2026Building Durable AI Agents46m
- 25 Jun 2026AIUC-1: Building trust in AI agents45m
- 11 Jun 2026Zero Trust for AI Agents47m
- 4 Jun 2026Breaking down the 2026 Stanford AI Index Report47m
- 28 May 2026Rebooting Enterprise AI with MCP and Kubernetes48m
- 7 May 2026The Myth of Model Wars: Open vs Closed AI in 202642m
- 13 Feb 2026AI incidents, audits, and the limits of benchmarks42m
- 17 Dec 2025Beyond chatbots: Agents that tackle your SOPs45m
- 10 Dec 2025The AI engineer skills gap45m
- 10 Nov 2025Are we in an AI bubble?49m
- 16 Oct 2025Dealing with increasingly complicated agents54m
- 9 Oct 2025The impact of AI on the workforce: A state-level case study44m
- 11 Sep 2025Cracking the code of failed AI pilots46m
- 27 Aug 2025GenAI risks and global adoption43m
- 12 Aug 2025Confident, strategic AI leadership47m
- 1 Aug 2025Workforce dynamics in an AI-assisted world44m
- 7 Jul 2025AI in the shadows: From hallucinations to blackmail44m
- 14 Apr 2025Orchestrating agents, APIs, and MCP servers42m
- 15 May 2024Full-stack approach for effective AI agents47m
Daniel Whitenack says he frequently sees online ads for courses on becoming AI proficient, including from Harvard.
“I'm all this all the time seeing these like ads for whatever it is. Uh, Harvard school, uh, like, what the engineering or what? uh, uh, becoming AI proficient in your business”
Daniel Whitenack says Prediction Guard sponsors the Midwest AI Summit, scheduled for October 15 in Indianapolis.
“We're also involved and a sponsor of the Midwest AI Summit, which is coming up October 15th in Indianapolis.”
Daniel Whitenack describes Prediction Guard as an AI control plane that customers host themselves, including on-premises, hybrid, cloud or air-gapped setups.
“Prediction Guard is an AI control plane that you host within your own environment, whether that's on prem, hybrid or in your cloud VPC, even air gapped scenarios.”
Daniel Whitenack says his company built a team of digital agents with an agent harness to help an organization's translators, with results appearing in the translation management system they already use.
“we have essentially a team of digital agents that helps translate with a with an agent harness.”
Daniel Whitenack is CEO of Prediction Guard and co-hosts the Practical AI podcast with Chris Benson, a principal AI and autonomy research engineer.
“I am CEO at Prediction Guard, and I'm joined as always by my co-host Chris Benson, who is a principal AI and autonomy research engineer.”
Daniel Whitenack gives an object recognition model as an example of an AI model that transforms an input image into the objects it contains.
“I give you an image and you tell me what's in the image, it transforms to objects in the image.”
Daniel Whitenack says an open model such as Gemma can be downloaded as weights from Hugging Face and run with software like the Transformers library.
“You can go to Hugging Face, you can download the weights of that, and there is software that runs it”
Daniel Whitenack says Nvidia and others had recently made public appeals about the importance of open-weight and open-source models.
“Nvidia and others making pleas to the community around the importance of open weight models or open source models.”
Daniel Whitenack says an agent is distinguished from an AI feature by being software meant to accomplish a goal, often with autonomy.
“What the software is meant to do is to accomplish a goal, and in many cases, to operate with autonomy.”
Daniel Whitenack says Formula One pit stops are a common leadership example because they improved from about 20 seconds to about 2 seconds.
“you'd look at how the pit stop evolved over time from like, whatever 20 s down to 2 s.”
Daniel Whitenack names LangGraph, Pydantic agents, Hermes agent and OpenClaw as examples of agent frameworks where the underlying model can be swapped.
“for example, LangGraph or Pydantic agents, or Hermes agent, or Open Claw, et cetera, et cetera.”
Daniel Whitenack cites Anthropic, OpenAI, IBM and AWS as vendors offering opinionated, vertically integrated agent stacks.
“inbuilt agent features in Anthropic or OpenAI's platform, but also others, whether it's like IBM Watson agent thingy”
Daniel Whitenack says businesses on Google's productivity suite get Gemini built in, while Microsoft shops get Copilot, and neither fits as a full digital workforce.
“if you're primarily using Google Drive and Google Docs, then you're automatically gonna get one of these very opinionated AI integrations across those various apps.”
Daniel Whitenack said news highlights reported that IBM's stock plunged about 25 percent.
“IBM had a major hit in their stock, so 25 stock plunge, which is is kind of crazy.”
Daniel Whitenack said IBM's single-session collapse erased about $70 billion in market value.
“25 single trading session collapse, which is about 70 billion in market value.”
Daniel Whitenack said IBM's one-day drop was its worst in over 50 years, exceeding its losses in the 1987 Black Monday crash.
“Marking it the, uh, worst single day drop in over 50 years, um, outpacing even its losses during the 1987 Black Monday market crash.”
Daniel Whitenack said analysts cite companies moving technology budgets away from enterprise software and services as a dynamic behind IBM's fall.
“companies are diverting their IT budget or their technology budget away from enterprise software and and services”
Daniel Whitenack said Anthropic is focusing more on implementation and services, as OpenAI has also done, rather than only on models.
“Anthropic really doubling down on kind of implementation and services side, which we've also seen with with Openai versus kind of the model side.”
Daniel Whitenack described a Chinese token black market in which brokers resell discounted API access to Western AI platforms like OpenAI and Anthropic, which are blocked in mainland China.
“brokers are reselling discounted API access to Western artificial intelligence platforms. AI platforms like Openai or Anthropic”
Daniel Whitenack said platforms such as Unsloth let people fine-tune models even on a MacBook.
“there's platforms like Unsloth and others that allow you to fine tune, uh, you know, even on your MacBook”
Daniel Whitenack said Prediction Guard recently spoke with a company running tens of thousands of agents, reported in the transcript as '70 zero', and another with a somewhat smaller number.
“we've talked to one one company that already has 70 zero agents running, one that has, I think it was like six zero or something.”
Daniel Whitenack said some enterprise software vendors like NetSuite provide connectors into the AI world, often via MCP, on the assumption that their value lies in their data platform.
“Which is often MCP, right? So their value is, maybe they're assuming their value is in their data platform, how they organize data”
Daniel Whitenack said that on the day of IBM's drop Workday, Salesforce, ServiceNow and Adobe fell by about 10, 9, 8 and 6 respectively, figures he found through an AI agent's research.
“Workday Workday slid down ten, Salesforce nine, ServiceNow eight, Adobe six, um, all, you know, uh, s similar dynamics going on.”
Daniel Whitenack said Nvidia and Intel rose that day as buyers rushed to purchase hardware.
“What did gain was, uh, uh, the, you know, Nvidia, Intel chip chip providers rising as people again, you know, panic, buy some of these, uh, these hardware solutions.”
Daniel Whitenack said Anthropic recently released a paper on verbalizable representations forming a global workspace in language models.
“Anthropic J did just release a paper. Um, I I forget which day re, uh, recently now, whenever it was, uh. Verbalizable representations from a global workspace and language models.”
Daniel Whitenack read the Anthropic paper as showing models have an improving functional ability to route and report information, without claiming they have feelings.
“there is this functional capability of these models to an improving functional capability of these models. To route and report information.”
Anthropic released its 'Zero Trust for AI Agents' framework around May 27, 2026, according to Daniel Whitenack.
“they on May, I think, twenty seventh of this year, 2026 released this. I guess it's a ebook, white paper, blog post, however you want to frame it, a framework”
Anthropic also released a product called Claude Security, which Daniel Whitenack describes as using AI for security rather than securing AI.
“They also recently released Cloud Security, which is more on the AI for security side, not so much the security for AI side.”
Zero trust security assumes threats are already inside the network and treats every user, device and request as a potential threat.
“A zero trust approach to cybersecurity, on the other hand, would actually assume that e everything inside the network, uh, uh, that that threats are already inside your network”
Daniel Whitenack estimates that more than ninety percent of enterprises with AI deployments do not operate under Anthropic's zero trust model.
“probably ninety of plus of of organizations, enterprises that have AI deployments currently are not operating according to this model.”
Least agency is a term attributed to OWASP that extends the principle of least privilege to agentic applications.
“least agency, which I guess is a term coined by OWASP, and and that extends this kind of idea of least privilege to agentic applications.”
Blast radius measures the potential damage if an AI agent goes wrong.
“It measures the potential damage if something goes wrong, if an agent does does go off the rails of that blast radius.”
Whitenack described writing a PDF interview exercise with extra hidden text intended to make Claude Code do the opposite of the visible instructions, as a prompt injection test.
“I had all the instructions in black text, and then I had an extra like three fourths of a page.”
An agent given limited instructions about a FastAPI server could discover other routes through its Swagger documentation at the /docs endpoint, according to Whitenack.
“It could just look at the Swagger documentation at the Slash docs endpoint and know about all the other routes that maybe it shouldn't use.”
Whitenack, summarizing the framework, says agents often run with elevated privileges or service accounts that human-oriented identity systems handle poorly.
“Agents often operate with elevated privileges or service accounts, and traditional identity systems designed for humans struggle to accommodate them.”
Anthropic's framework defines three tiers of controls: foundation, enterprise and advanced.
“They present three capability. W- I think what they call capability tiers, or three tiers of application.”
At the foundation tier of identity verification, the framework calls for unique cryptographic identifiers for each agent instance.
“There is to have unique cryptographic identifiers for each agent instance, so to assign persistent agent IDs backed by cryptographic material, not just labels.”
The framework's enterprise tier uses certificate-based authentication, and its advanced tier uses hardware-backed identity with attestation.
“The enterprise level is certificate based authentication with full life cycle management, and the advanced is hardware backed identity with attestation”
The framework's foundation tier for access control is role-based access control with deny by default.
“Access control Foundation level is role based access control, or RBAC with deny by default, that's the the foundation in in that category.”