SHOW: 927

SHOW TRANSCRIPT: The Cloudcast #927 Transcript

SHOW VIDEO: https://youtube.com/@TheCloudcastNET

CLOUD NEWS OF THE WEEK - http://bit.ly/cloudcast-cnotw

NEW TO CLOUD? CHECK OUT OUR OTHER PODCAST - "CLOUDCAST BASICS"

SPONSORS:

• [VASION] Vasion Print eliminates the need for print servers by enabling secure, cloud-based printing from any device, anywhere. Get a custom demo to see the difference for yourself.
• [US CLOUD] Cut Enterprise IT Support Costs by 30-50% with US Cloud

SHOW NOTES:

• Patronus AI website

Topic 1 - Welcome to the show, Anand. Give everyone a quick introduction.

Topic 2 - Our topic today is Preventing AI Model Hallucinations. Before we dig into that, I wanted to ask about your time as Lead Data Scientist at Meta. What was it like to be early into that organization, and what did you take away from your time there?

Topic 3 - Ok, let’s dig into model evaluations and hallucinations. Let’s start at the beginning. How do model hallucinations come about?

Topic 4 - When evaluating models for hallucinations, how does a developer or a data scientist know fact from fiction? Due to its size, complexity, and number of parameters, it’s not feasible to simply fact-check and manually verify inputs to outputs. How is this process evaluated and automated with some level of confidence? Additionally, numerous benchmarks are available. What are your thoughts on the usefulness of the benchmarks?

Topic 5 - How does the concept of data quality play into this? How would we know when a model was given insufficient or improper data vs. a hallucination

Topic 6 - We often hear about how frontier models are running out of training data, and increasingly, synthetic data is being used. Does this impact hallucinations in any way?

Topic 7 - The last item I wanted to ask you about, Partonus AI, also pertains to model optimization. Can you explain that process?

Topic 8 - If anyone is interested, what’s the best way to get started?

FEEDBACK?

• Email: show at the cloudcast dot net
• Bluesky: @cloudcastpod.bsky.social
• Twitter/X: @cloudcastpod
• Instagram: @cloudcastpod
• TikTok: @cloudcastpod

The industry is debating whether AI should be viewed as augmenting or replacing human-centric tasks. But we don’t yet have a framework to discuss the technical and business impacts of that spectrum of decisions.

SHOW: 926

SHOW TRANSCRIPT: The Cloudcast #926 Transcript

SHOW VIDEO: https://youtube.com/@TheCloudcastNET

CLOUD NEWS OF THE WEEK: http://bit.ly/cloudcast-cnotw

CHECK OUT OUR NEW PODCAST: "CLOUDCAST BASICS"

SHOW SPONSORS:

• [US CLOUD] Cut Enterprise IT Support Costs by 30-50% with US Cloud
• [VASION] Vasion Print eliminates the need for print servers by enabling secure, cloud-based printing from any device, anywhere. Get a custom demo to see the difference for yourself.

SHOW NOTES:

• The final 5% of AI success

HOW DO WE THINK ABOUT THE INVOLVEMENT OF HUMANS WITH AI SYSTEMS?

• What do we think AI systems should be capable of?
• What do we think human systems should be capable of?
• We know how to quantify error, but do we really know how to quantify error?
• Almost every question about AI comes down to augment vs. replace, and yet humans tend to skew towards the replace angle.
• We like the thought of humans in the loop, but often don’t want to be bothered to interact with humans
• We haven’t yet created a framework to think about the human +/- AI involvement, because we haven’t really created it for general-purpose automation yet either.

FEEDBACK?

• Email: show at the cloudcast dot net
• Twitter/X: @cloudcastpod
• BlueSky: @cloudcastpod.bsky.social
• Instagram: @cloudcastpod
• TikTok: @cloudcastpod

Joel Christner, (@joelchristner, Founder/CEO at @viewyourdata) discusses the complexities of data management in AI, structured and unstructured data, the importance of RAG pipelines and vector databases.

SHOW SUMMARY: Aaron and Joel discusses the complexities of data management in AI, focusing on the concept of universal data representation. They explore the challenges organizations face with structured and unstructured data, the importance of RAG pipelines and vector databases, and the implications of data privacy in regulated industries. The conversation also touches on managing model versions and the emerging patterns in AI tooling that can help enterprises effectively utilize AI technologies.

SHOW: 925

SHOW TRANSCRIPT: The Cloudcast #925 Transcript

SHOW VIDEO: https://youtube.com/@TheCloudcastNET

CLOUD NEWS OF THE WEEK - http://bit.ly/cloudcast-cnotw

NEW TO CLOUD? CHECK OUT OUR OTHER PODCAST - "CLOUDCAST BASICS"

SPONSORS:

• [VASION] Vasion Print eliminates the need for print servers by enabling secure, cloud-based printing from any device, anywhere. Get a custom demo to see the difference for yourself.
• [US CLOUD] Cut Enterprise IT Support Costs by 30-50% with US Cloud

SHOW NOTES:

• View.io website

Topic 1 - Welcome to the show, Joel. Give everyone a quick introduction.

Topic 2 - Our topic today is everything data and how to represent it and embed it into AI systems. First, what is the challenge with data, structured or unstructured, in organizations today and what is behind the concept of Universal Data Representation

Topic 3 - Industry or customer specific data today is big challenge for organziations, especially in highly regulated industries such as healthcare, financial services, etc. The most prevalent solution I am seeing is taking an existing foundational model and then adding a RAG pipeline vs. the cost and time to fine tuning. What are you seeing?

Topic 4 - Even when companies have good data, that doesn’t mean that data makes it into the AI pipeline correctly, this is where the embedding problem and your concept of Universal Data Representation comes into play, correct?

Topic 5 - But, once you get the first model out, then what? How should the data and models be handled over time? How do you create a platform and a continuous feedback loop to improve the results over time?

Topic 6 - What are the most successful use cases you are seeing today with your customers?

FEEDBACK?

• Email: show at the cloudcast dot net
• Bluesky: @cloudcastpod.bsky.social
• Twitter/X: @cloudcastpod
• Instagram: @cloudcastpod
• TikTok: @cloudcastpod

Where does the next phase of AI-assistants for software development go next? Is it an evolution of developer productivity, or a complete rethinking of the barriers and limitations for broader software development?

SHOW: 924

SHOW TRANSCRIPT: The Cloudcast #924 Transcript

SHOW VIDEO: https://youtube.com/@TheCloudcastNET

CLOUD NEWS OF THE WEEK: http://bit.ly/cloudcast-cnotw

CHECK OUT OUR NEW PODCAST: "CLOUDCAST BASICS"

SHOW SPONSORS:

• Cut Enterprise IT Support Costs by 30-50% with US Cloud

SHOW NOTES:

WHERE DO AI DEVELOPER-ASSISTANTS GO NEXT?

• A year ago it felt like co-pilots were one of the entry point use-cases for AI.
• Since then we’ve seen numbers say the uplift is 10-20% productivity.
• Microsoft claims that 20-30% of their code is now written by AIs.
• We’ve seen many senior developers speakout that it’s not replacement level technology and they don’t trust it.
• Companies like Cursor have a $9-10B valuation. Windsurf just got purchased for $3B by OpenAI.
• Microsoft has Co-Pilot (based on OpenAI models). Google and Amazon are rumored to be launching their own.
• Lots of companies are launching agents (IBM, Salesforce, Oracle, etc.), and lots of agent frameworks now exist.
• So where does it go next? Is it just wide-spead adoption of developer productivity?
  • Is it specialized functions within developer workflows? (e.g. CI/CD, documentation, security evaluations, bug fixes, long-term maintenance, etc.)
  • How far are we from teams being just a few architects, leads, Sr. Devs, and then teams of AI’s (agents, etc.)?
    • Is that a good thing for Sr. Dev personalities that didn’t want to focus on soft-skills?
    • Does that allow for greater experimentation against feature-requests or stories, since they can create more, test more, etc.?
    • Do we start to see companies create skunkworks teams/groups that try to adopt this approach?
    • Are product managers ready to have zero-backlogs and the demand for new ideas to increase?

FEEDBACK?

• Email: show at the cloudcast dot net
• Twitter/X: @cloudcastpod
• BlueSky: @cloudcastpod.bsky.social
• Instagram: @cloudcastpod
• TikTok: @cloudcastpod

Siqi Chen (@blader, CEO/CFO @Runwayco), talks about his journey from JPL developer to Founder of a financial planning and analysis (FP&A) startup. We focus on how to build products that customers crave and how a customer-centric view differs from traditional product management.

SHOW: 923

SHOW TRANSCRIPT: The Cloudcast #923 Transcript

SHOW VIDEO: https://youtube.com/@TheCloudcastNET

CLOUD NEWS OF THE WEEK: http://bit.ly/cloudcast-cnotw

NEW TO CLOUD? CHECK OUT OUR OTHER PODCAST: "CLOUDCAST BASICS"

SPONSORS:

• Cut Enterprise IT Support Costs by 30-50% with US Cloud

SHOW NOTES:

• Runway website
• Behind What Seems Like an Overnight Success (video)

Topic 1 - Welcome to the show, Siqi. First, your combination of technical and business/financial background is fascinating. How did you go from coding at NASA to Head of Product at Zynga to CEO/CFO for a finance platform startup? Give everyone a quick introduction.

Topic 2 - One thing I’ve noticed as a trend in your background is the core concept of building. What has been your philosophy in building products? How do you build products that customers demand?

Topic 3 - Let’s talk about AI and AGI for a moment. We hear all the time how disruptive this will be. What are your thoughts here, and how do we develop both adaptability and resiliency to new technologies?

Topic 4 - Let’s talk FP&A (financial planning & analysis). Our core listeners out there tend to skew more towards the tech and infrastructure side, but a core theme of this show is always to be learning as much of the business as possible to apply those concepts. As someone with a background in both worlds, plus now running an FP&A startup, what do you wish folks on the technical side of the house knew more about to make their jobs easier?

Topic 5 - We posted a link in the show notes for a video you did on the “overnight success” of Runway. It was a good representation and origin story of how something can go viral with the right mindset and product-market fit. Tell everyone about that as Runway approaches 5 years now.

Topic 6 - What is your biggest challenge in the FP&A space today? Is it AI? We’ve seen a lot of AI disruption in coding, legal, and other areas requiring deep data pool insights. Is this any different?

FEEDBACK?

• Email: show at the cloudcast dot net
• Bluesky: @cloudcastpod.bsky.social
• Twitter/X: @cloudcastpod
• Instagram: @cloudcastpod
• TikTok: @cloudcastpod

Tech CEOs are making bold proclamations, from WhyQ to It is what it is. How will companies navigate this spectrum as they seek innovation, accountability and profitability throughout 2025?

SHOW: 922

SHOW TRANSCRIPT: The Cloudcast #922 Transcript

SHOW VIDEO: https://youtube.com/@TheCloudcastNET

CLOUD NEWS OF THE WEEK: http://bit.ly/cloudcast-cnotw

CHECK OUT OUR NEW PODCAST: "CLOUDCAST BASICS"

SHOW SPONSORS:

• Cut Enterprise IT Support Costs by 30-50% with US Cloud

SHOW NOTES:

• Amazon 2024 CEO Letter to Shareholders - A Why Culture
• Uber CEO says “it is what it is” about changing benefits

WHEN DOES WHY BECOME IT IS WHAT IT IS?

• Communicating is difficult, especially as the company grows in size (or is remote)
• Communicating change is difficult, even when communication channels are strong
• Outside of finance, tech tends to pay at the high end of salaries and perks
• We’re in an interesting time of challenging economics and pressure from AI
• Perks are difficult to pull back, because business success isn’t evenly rewarded
• Why is positioned as open culture, or strategy, but it’s also about day-to-day behavior
• It is what it is a decision, but it’s also accountability and continued viability
• Most leaders are going to have to manage between Why and It is what it is in 2025
• Most workers are going to have to work between Why and It is what it is in 2025

FEEDBACK?

• Email: show at the cloudcast dot net
• Twitter/X: @cloudcastpod
• BlueSky: @cloudcastpod.bsky.social
• Instagram: @cloudcastpod
• TikTok: @cloudcastpod

Aaron Delp, Brian Gracely, and Brandon Whichard discuss the top stories in Cloud and AI from April 2025, including OpenAI, MCP, VibeCoding and Hyperscaler earnings.

SHOW: 921

SHOW TRANSCRIPT: The Cloudcast #921 Transcript

SHOW VIDEO: https://youtube.com/@TheCloudcastNET

CLOUD NEWS OF THE WEEK: http://bit.ly/cloudcast-cnotw

NEW TO CLOUD? CHECK OUT OUR OTHER PODCAST: "CLOUDCAST BASICS"

SPONSORS:

• Cut Enterprise IT Support Costs by 30-50% with US Cloud

SHOW NOTES:

• Link to April 2025 News and Articles

FEEDBACK?

• Email: show at the cloudcast dot net
• Bluesky: @cloudcastpod.bsky.social
• Twitter/X: @cloudcastpod
• Instagram: @cloudcastpod
• TikTok: @cloudcastpod

As more companies begin to adopt AI into their workforce and day-to-day processes, it will be interesting to watch how their learning curve is spread across knowledge workers.

SHOW: 920

SHOW TRANSCRIPT: The Cloudcast #920 Transcript

SHOW VIDEO: https://youtube.com/@TheCloudcastNET

CLOUD NEWS OF THE WEEK: http://bit.ly/cloudcast-cnotw

CHECK OUT OUR NEW PODCAST: "CLOUDCAST BASICS"

SHOW SPONSORS:

• Cut Enterprise IT Support Costs by 30-50% with US Cloud

SHOW NOTES:

• AI Horseless Carriages (AI user-experiences)

HOW WILL WE VIEW AN AI AGENT IN THE CONTEXT OF HUMANS OR “USERS”

• The low-hanging fruit, simple on-ramp is the key to early AI adoption
• Google and Microsoft are already showing revenue increases, likely through the productivity apps bundling
• Expect prices to increase slowly, but frequently as adoption happens and companies get used to the knowledge worker productivity increases (or expectations)
• Curious how knowledge workers are adopting, sharing, increasing their learning curve
• Sharing still seems to be lacking within the AI tools. Not just sharing of an individual task, but sharing of learning curves, best practices, datasets
• Is there a dataset collection opportunity? This feels like Big Data or Data Lake 5.0.

FEEDBACK?

• Email: show at the cloudcast dot net
• Twitter/X: @cloudcastpod
• BlueSky: @cloudcastpod.bsky.social
• Instagram: @cloudcastpod
• TikTok: @cloudcastpod