Discover more from Coté's Wunderkammer
If it’s bullshit work, have the bullshit-artist do it - Using AI for Bureaucratic Toil Removal
I develop a new talk. Also, two predictions for 2023 (DevX and layoffs), the timelessness of airports, and more links.
Hey there! As I type this, SpringOne is being livestreamed right now! Chances are, you’ll read this when it’s over. But, you can go back and see today’s talks, and there’ll be more tomorrow and Thursday. If you’re interested in development, cloud native, and especially the question “what is VMware Tanzu?” today’s talks are for you. You can watch free, check it out and schedule it in for tomorrow too!
Also, don’t forget that if you are in or can make it to Amsterdam tomorrow, I’m hosting a watch party for day two. We’ll start with my recap of some of today. There’ll be drinks and food and some demos, Q&A. Come check it out, for free.
If it’s BS work, have the BS-artist do it - Using AI for Bureaucratic Toil Removal
I’m rushing to get to making dinner and family stuff, so I haven’t really edited the below at all. WASTEBOOKWUNDERKAMMERPOWER!!1!
I put the final touches on a new talk idea today. As you know, I’ve been trying to figure out how to use the AI (you know, ChatGPT) to make office work less shitty - removing white collar toil, so to speak. Biking home after dropping the kids off, I put through an outline in my head, and then messed around with it in the lab of live-streaming.
The idea of the talk is to come with three or so examples (demos!) of using ChatGPT (or whatever) to automate toil (to some, bullshit) at work.
Part of the gimmick of this talk is that I’d do some “live prompting,” even taking some ideas from the audience. Most of my peers get to live code and (you better sit down for this one least you fall over out of excitement) live configure kubernetes. I don’t get to do any of that. So, I’m hoping “live prompting” will be a cheap-trick to get to that thrill.
TK: Ethics and Aesthetics
Now, big “don’t @ me, bro!” upfront. I haven’t through through the ethics and aesthetics parts yet. Well, I have, but I haven’t gotten to that. I’m thinking of explaining Jevon’s Paradox but also my dad’s job history (getting sold off by IBM to another company, etc.). Jevon’s Paradox is benefits us all, but rarely benefits the automated…or something.
Aesthetic wise, some people might be repelled by some of the things you’d automate. Like writing a performance review or weekly status reports. I dunno, I’ll figure that stuff out.
So, here’s the things I tried to "live prompt:
This meeting could have been an email
Summarize a quarterly, all hands meeting. Like you, I’ve watched a lot of quarterly meetings (QBRs, etc.). I crave the information in them, but they can be…uh…like a buffet that you requires you to eat one of everything. I wanted to take a transcript from an all hands meeting and get it summarized.
It didn’t really work too well, but I think that’s because I free-styled a meeting - like, I just made one up with voice-to-text as you can see. I need some actual transcripts, but these kinds of meetings are not really for the public. I was thinking maybe I could take some transcripts from open source SIGs, or some kind of recording of “how we’ve been doing” session for an open source foundation or project.
Argue for process change
If you’re doing (pardon me) digital transformation, a lot of the work you’re doing is convincing people to change how they work. This is not only a lot of emails, but talks, and just casual conversations. I picked pair programming as a topic. I figure this is something most people will think is crazy and you’d have to spend a lot of time on. This one took several rounds, but after discovering that you can pass in PDFs and web pages to it, ChatGPT started putting together a good argument. This one works!
Brainstorm vision and strategy
One of the major things an executive (or any “leader”) needs to do is come up with a vision that can be used as a tool. I talk about this a lot. Here, ChatGPT was half successful. I think it took my ideas and came up with a good business vision, and even the outcomes you’d want to get to that vision. But when it came to telling developers what they’d do, it went too generic. I think if I had a of ideas, it would have done better. But…then I might should have just written it myself. I should trying cleaning up the transcript of me talking about this and see if it can use it to come up with a more formal thing.
Writing a FAQ on new policies in your new culture
You’re doing the digital transformation and introducing all sorts of new policies and practices. At scale (thousands and thousands of staff), you’re going to need to do a lot of internal comms, FAQs and stuff. So, let’s say you switch to unlimited PTO. Let’s have ChatGPT start a FAQ for us.
I picked this because it’s well documented and is representative (enough) of a new policy that would match being a kinder, more trusting organization - shifting right on Westrum’s columns. This is also a good example because there’s a lot of controversy around unlimited PTO from the DevOps, tech-world, etc. people.
In this case, I fed some research and with a little bit of work, I think it did a great job: see the results. First, you’d need to clean it up to a final email, but you some good parts there. And, I mean, yeah, you’d want to take this and play with it more, run it by HR, etc. But this would save a ton of bootstrapping time, and refining.
Each of the above is text, emails and FAQs. My intention here is that this stuff would also be useful for verbal stuff, face to face, and as with the vision stuff, just thinking through it. But, yeah, lots of text.
I need to spend a lot more time learning the interface and how it all works. Leading directly to:
In most of these cases, I could have written the answers faster than it took me to get something good out of ChatGPT. However, my theory is that I just need to learn better prompting. For example with the pair programming one, if I’d known you could pass in research papers via URL, I think I’d have gotten to it much faster.
I didn’t spend time fact checking many of the things, but you’d need to based on some of the results.
Ironically, getting more humans to work on this, adding in their own ideas for prompts and tuning, would probably make it better and faster.
Thanks for reading Coté's Wunderkammer! Subscribe for free to receive new posts and support my work.
If it’s BS work, have the BS-artist do it - Using AI for Bureaucratic Toil Removal
I think I validated the idea for a talk, so I wrote an abstract, below. I submitted it to one conference today (DevOpsDays Austin) and I’ll get to submitting it to more soon.
There’s a lot of bureaucratic toil you need to deal with when you’re improving your software culture, tools, and methods inside large organizations. You know: DevOps, digital transformation, and stuff. There’s also just the daily corporate nonsense you need to deal with. This talk will explore using AIs to help reduce bureaucratic toil like: keeping up with status meetings and weekly reports, dealing with glacial-speed governance, persuading people to embrace new ideas like pair programming, vision and strategy, writings tories and managing backlogs, transforming HR policy, working with finance, dealing with the ticket-wall, and more. You’ll see “live prompting” with ChatGPT, even from audience suggestions. Come experiment with using AI to cut through the kind of toil that’s keeping your organization in Westrum’s Bureaucratic column.
Note to conference organizers
DevOps and SRE taught us to automate as much toil as possible. I think we can use things like ChatGPT to automate bureaucracy toil. This talk will do “live prompting” to explore that theory and leave the audience with some tools to spend less time on corporate BS.
Why I’m excited about this talk
I don’t like dealing with meetings, writing reports, trying to persuade people to change to proven techniques…all that stuff those meetings that could have been a meeting, or just a one line Slack message. No one does! I’ve been playing around with ChatGPT to write this kind of stuff, but also do things like brainstorm ideas for things like vision and strategy. I’m most excited to get feedback from the audience and in doing some impromptu “live prompting” find out even more and…have fun.
VMware 2023 Predictions: Platform Engineering Improves Developer Experience, Tech Layoffs Solve Enterprise Talent Gaps - I make two predictions/comments: (1) we’re probably done with tinkering around kubernetes and can now move up to the stack to focus on developer tools (2) there’s a lot of tech layoffs, so normie companies should have some good people to hire.
Relevant to your interests
Banks closed 93 percent of branches since 2001 - “Many bank branches closed in the past few years. At the beginning of 2020, Rabobank still counted 335 branches. It now has 136. ING went from 170 bank branches to 57 now. ABN Amro only has 27 bank branches in the country. You can only find one ABN Amro branch in six of the twelve Dutch provinces.” // Weird compared to the States…? Also: I went to one of my bank branches (ABN) a year or so ago to withdrawal some money. They were like “oh, we don’t do that anymore.” So I had to go find an ATM that had a large enough withdrawal limit to get what I wanted. But, I mean: wut?
Dumb and shameful until it’s not - “This is why there shouldn’t be journalists-only instances. Journalism is neither a real community or an identity. It’s a job and if you insist on making it your identity the only central point most seem to land on is view from nowhere bullshit that breaks your critical thinking.”
Embracing the Evolution to Cloud-Native Infrastructure - An impressive attempt to jam a cloud native tech definition into one paragraph: “Compared with machine-centric virtual infrastructure, cloud-native infrastructure is fundamentally application-centric. When based on Kubernetes, cloud-native infrastructure introduces some practical changes, such as pods effectively becoming the CPUs; persistent volume claims (PVCs) becoming the data storage devices, and service connectivity layers such as services meshes becoming the network. Cloud-native infrastructure will also exploit the evolution of compute, storage, and networking technology at lower levels of the infrastructure, such as running containers on bare metal servers; offloading tasks to specialized function accelerator cards (FACs); using processors based on architectures such as ARM, and running code with micro-VM approaches like WebAssembly (Wasm).”
Cloud Native Predictions 2023, from Chris Aniszczyk at the CNCF.
Coding for the future of U.S. national defense - VMware Tanzu in the US Military - “The Army had unsuccessfully spent 19 years with various contractors to create the mapping/placement software.The Army Software Factory did the job in just three months.” // These examples of how parts of the US military have revolutionized how they do software is a great case study. We’ve worked with them over the years starting at Pivotal and now VMware Tanzu. They’ve talked a lot over the years about what works and doesn’t and have great results. Part of the fun of my job is hearing these kinds of stories from all sorts of organizations and companies. It’s possible to change if you put in place the right approach and, most importantly, keep up the discipline of following it month after month, year after years.
US Startups And Tech Companies With Job Cuts In 2022 and 2023. - As of Jan 20th, 2023: “More than 46,000 workers in U.S.-based tech companies have been laid off in mass job cuts so far in 2023.” Also with ongoing spreadsheet. Some more commentary here on B2B company layoffs. And a theory that our old pals hedge funds/PE asked for people to get laid off. Real fun bunch, that lot.
Tipping Is Weird Now - The Atlantic - “Square sent me data showing that tips received by both full-service and quick-serve restaurants exploded from 2020 to 2021; growth continued in 2022, but more modestly—full-service was up by more than 25 percent in the third quarter of 2022, and quick-service restaurants were up nearly 17 percent. Despite complaints, people are still tipping well and often.”
Five Days in Class with ChatGPT - “Another, related benefit really stunned me: everybody could keep up. We had a highly uneven level of technical expertise in the classroom, from no technical background to computer science degrees. In the before-time we would have lost at least half the class by day three. “I would have gotten lost in class several times without ChatGPT,” said Martin Wendiggensen, one of the more technical students. Same for me.”
I have a life, so I don’t have time to keep up with real life.
“I don’t know how to explain it, but time works differently in airports. 5 minutes and 5 hours are somehow the same. Your flight is delayed but arrives early. The carpets are from 1980. The water machines are from 2045. It is always happy hour. You have always been in the airport.” -Tiffany Li
It’d be nice to have more images, right? But, I gotta get going. I’ve been sort of clogged up when it comes to new content - not really writer’s block, more like “giving a shit block” :) So, it’s nice to have finally cracked something new to work on. It takes a long time to get that point: a lot of staring into the distance, thinking it through. But, it’s nice to just sit down and see what’s there.
See y’all next time!