Interview: Brian Gracely

Whitney and I interviewed Brian Gracely for this week’s Software Defined Interviews episode. It was a great, big ol’ basket of topics: the process of gathering and reporting cloud news, the evolution of PaaS, and the pros and cons of working at small startups versus large companies. Also: career advice, the importance of communicating value within organizations, and how to stay relevant in the ever-changing tech landscape. And still more: Brian shares insights on how to generate engaging content for podcasts and the impact of internal communication on company culture.

If you don’t already subscribe to the podcast, you should do that! You can just go to the show page and listen to it there too, and there’s the video recording of the episode too.

What went wrong with PaaS?

In the interview, I asked Brian how PaaS went wrong in the 2010’s, and he gave a well thought out answer. I can’t say I disagree with him, especially in hindsight.

You should, of course, listen to his whole answer, but here’s what I’ve been thinking about since talking with him:

1. The PaaS vendors at the time started off focused on one language and, arguably, one buyer. Heroku was focused on startups and smaller development teams, specializing in Ruby. Cloud Foundry sort of started focusing on Java, but when it finally was released, it was very ruby focused. You can see a tension, almost, between those two languages in the first announcement.

2. When PaaSes tried to be “all things to all people,” it spread things too thin. It was hard to support all the little wing-dings and shiny objects each language community wanted.

3. This led to “opinionated platforms,” restricting what developers can do as a feature. For the record, I think this is incredibly valuable. Chasing “the shiny and new” is a silent killer for enterprise app strategy.

4. People wanted to customize, and was a window Kubernetes took advantage of. It promised that you could build you own PaaS and get it exactly how you wanted it.

5. Thus, the PaaS idea waned.

As you know, dear readers, I think this has been a setback, going on almost ten years now. Hopefully we’re finally emerging from the platform dark ages after the container orchestration calamity and can get back to focusing on platforms. Just remember to not destroy our progress it all in five or so years.

Wastebook

• He’s like if Martin Fowler and Hunter Thompson had a baby.

• “magical coherence,” on Apple.

• “I’m 64 years old. I’ve been an actor since I was 19. I made horror films and sold yogurt that makes you shit. I never thought I would hear my name at the Oscars.” Jamie Lee Curtis.

• “Discoursepolitik,” and:

• ”Invasion Of The Bluesky Disagree Bots.”

• There’s a passage in You Shouldn’t Have Come Here about eating beans and weenies that I’m pretty sure is metaphor for trepidation about having sex. (Also, it has bacon in it, which is a good twist on some classic, American trash-food.)

• ”One Hand To Shake Beats One Throat To Choke” Naveen Chhabra, Forrester. “common culture is a delusion of my age,” Warren Ellis.

• Are any of these big-ass spending announcements ever true? Also, even though their product is amazing, these OpenAI guys are seeming more and more sketchy.

Relative to your interests

• A free, powerful Chinese AI model just dropped — but don’t ask it about Tiananmen Square - Great, now the AIs are getting political. // I guess you could also so: at least, politics I don’t like.

• Deepseek: The Quiet Giant Leading China’s AI Race - ”Open source, publishing papers, in fact, do not cost us anything. For technical talent, having others follow your innovation gives a great sense of accomplishment. In fact, open source is more of a cultural behavior than a commercial one, and contributing to it earns us respect. There is also a cultural attraction for a company to do this.”

• Ignore the Grifters - AI Isn’t Going to Kill the Software Industry - “we’ll eventually allocate our scarce AI resources towards the things they are best at, which leaves plenty of things for humans to do.” // Among other things, there’s a simple point about technology advancements that threaten human employment: there is so much work to do that we can find more for everyone, machine and meat sack.

• 4 Lessons We Learned from Bringing AI to Our Company - “Next, you’ll face potential roadblocks, as privacy and security teams will be looking into where the models you use are hosted and where the data is stored. Chat.com, Gemini.com, or anything free of charge and a privacy nightmare is out of the question.”

• Why run AI on-premise? - Big ol' list of why you’d chose private cloud. // “While cloud-based AI services offer scalability and cost-effectiveness, especially for testing and early use cases, enterprises are increasingly considering on-premise AI solutions. Factors such as data sovereignty, security, performance, and cost are driving this shift, particularly as AI projects grow and require more data and processing power. Enterprises are also exploring less resource-intensive models and open-source options to reduce costs and improve efficiency.”

• OpenAI releases Operator agent as rivals enhance their AI services - ”When users ask Operator to perform a task in a website, the agent navigates to the relevant URL using a built-in browser. It can type, click and scroll to carry out the requested action. Operator regularly takes screenshots of the website to check that everything is working as expected. ” // Hopefully they’ll do some MCP here to be industry plugin friendly, but it’s doubtful.

• Are better models better? - “Part of the concept of ‘Disruption’ is that important new technologies tend to be bad at the things that matter to the previous generation of technology, but they do something else important instead.”

• How to use NotebookLM Plus for your business - The fact that most AIs can’t generate PowerPoint might be great: people will stop using decks for operations and go to the more helpful document. Slides for daily operations is terrible way to run the railroad. You could start doing PRFAQs and 6 pagers.

• Intuit CTO keeps 8,000 engineers on track with a base platform of common services - Centralizing, standardizing.

• The 2024 State of Platform Engineering? Fledgling at Best - “The majority of the organizations surveyed – 56% – have had platform teams for less than two years. A mere 13% of respondents reported working in ‘platform engineering’ for more than five years.”

• Observability: the present and future, with Charity Majors - “The main trend across the industry: consolidation. Companies try to control their bills.”

• Return to PaaS: Building the Platform of Our Dreams - It’s a tough slog.

• RTO Watch: “75% of workers with jobs that could be done remotely said their employer has put in-person mandates in place, according to a Pew Research survey conducted last fall and released last week. That’s up from 63% in 2023.” // RTO vs. WFH is a management vs. workers issue. I’m convinced that RTO has no rational basis, it’s just a culture that’s chosen or not. This could be great, this could be bad, but it’s a choice of how you structure the system and the negatives you accept (you limit your labor supply).

• Empowering Your IT Future: The Shift to Private Cloud - (1) Cost control vs public cloud (2) Data sovereignty (especially in APAC and EU) (3) Private AI capabilities (4) Enhanced license portability across environments.

• M&A and the Product Model - Silicon Valley Product Group : Silicon Valley Product Group - At tech companies, achieving synergies is very difficult, and predicting if you can do the work is even harder. // “Most people experienced in due diligence have a reasonable understanding of assessing the product – the customers, the financials, the offering, the technology used to build that product, and especially the go-to-market for that product. Yet these same people often do not have the experience to understand what it will take to integrate the newly acquired product with the parent company’s existing systems and operations, or migrate the customers to the existing company’s systems, or in many cases, especially in cases of significant technical debt, to re-platform the product, or to change the product to work effectively with the parent company’s go-to-market channels.” // Keeping the acquired company as a mostly separate entity - product wise - is a good strategy, e.g., YouTube, maybe Red Hat? It means you won’t kill off the thing you acquired the company for by meddling with it, integrating it into the way your company works.

• The Product Model Solves For Tech Debt - ”Forrester does not recommend ROI as a criteria for deciding to rectify technical debt, which should be seen more as essential maintenance spend.” // If you ran rail company they shouldn’t spend money on train maintenance, they’d tell you would kill the business (and people). Software is the same, executives just need to get that through the head. The ROI of reducing and limiting of limiting tech debt is the ability to function after 12 to 24 months.

• Notable Sections of the 2024 D&D Dungeon Master’s Guide - I’ve been wondering if it’s worth reading the new DM guide. It looks like it.

• Spotify’s playlists have altered the music industry in unexpected ways - artists are often disrupted first. Study how their business models - and personal compensation changes - and you can prepare and defend yourself.

• The return of skinny jeans? Men’s catwalks suggest wide-legged trousers are out and calf-huggers back in - Looks like a dodged a bullet on this one by wearing my old Gen-X pants despite the trends.

• Alternatives to “sorry” - American English meanings for “sorry.” // Yes, and:: in the Netherlands (and other parts of Europe), “sorry” is usually used like the en_US “excuse me” or “pardon me.” If you’re walking into a busy tram and you need to press through people you say “sorry, sorry.” And, it’s the English word everywhere not the local equivalent.

Conferences

Events I’ll either be speaking at or just attending.

cfgmgmtcamp, Ghent, Feb 3rd to 5th, speaking. SREday London, March 27th to 28th, speaking. Monki Gras, London, March 27th to 28th, speaking. CF Day US, Palo Alto, CA, May 14th. NDC Oslo, May 21st to 23rd, speaking. KubeCon EU, April 1st to 4th, London.

Discounts: 10% off SREDay London with the code LDN10.

Logoff

Also, there’s last week’s Software Defined Talk (lots of beans in chili talk, plus the usual nerd-shit), and here’s a thread of mine of reasons to be optimistic about DeepSeek lowering the cost of AI, or skeptical of freaking out. // If you’re into D&D, I started a tumblr that’s a commonplace book for my D&D thinking. That means it’s inspiration, scraps, concepts, and other things that get me thinking about and developing D&D ideas. I’m putting a lot of my AI generated stuff in there too, and maps I make. Obviously, it has plenty of content from others as I scroll and find it. Check it out!