Forrester’s Impressions: VMware Explore 2023 - A brief overview of VMware’s strategy and where Forrester thinks it going.

It’s Not You, It’s Me: What It Really Means When Budget Is The Reason For The Breakup - Something that’s not useful isn’t worth paying for.

The Works of Mars, 1671 - Fortification engineering.

Dangerous Dimensions: Mind-Bending Tales of the Mathematical Weird - Looks like fun, right?

Here’s Something Past Its Expiration Date: the Expiration Date Itself - “Food experts broadly agree that the expiration dates on every box of crackers, can of beans and bag of apples waste money, squander perfectly good food, needlessly clog landfills, spew methane and contribute to climate change.” // And, they’re gone for the most part in the UK.

Texas’s Biggest Barbecue City Is Attracting a New Crop of Exciting Restaurants - Lots going on in Lockhart.

Favorite coffee-making setups from the Ars Technica staff - I made coffee with a Chemex for a few years. The coffee was good, and the overall ritual of it was just fantastic.

Books Recommended With Uncommon Wisdom and Tender Care - “Over and over, Aoyama demonstrates how it’s done. In her Hatori ward, good fortune is not arbitrary or unearned; it is never a gauzy gift from the universe. It arises instead from action, experience and wisdom. Her characters appreciate each other; they are grateful to each other; they recognize in each other quality and potential. (Put these folks in a laboratory dish with the dramatis personae of a cynical HBO show and they’d annihilate each other, matter and antimatter.)” And: “You’ve got to be careful with novels about libraries and bookstores… The risk, in all these cases, is flattery. It feels nice to be assured that the places you find appealing are, in fact, wonderful. It’s also boring. The standard for such novels, therefore, is that they reveal something interesting and true about these environments."

Iowa

I’m in Des Moines, Iowa for the keynote I’ll be giving at DevOpsDays here. I believe their registration is closed, so…if you’re not already registered, I won’t be seeing you! But, you can check me out rehearsing it here. There’s 20 minutes of bonus content! Like and subscribe, MOFOS!

ChatDM - Dealing with short-term memory

If you’re interested in my recent experiments to get ChatGPT to act a Dungeon Master for solo-play Dungeons & Dragons, check out my recent write-up. It’s going OK, and I think I’ve figured out a technique to get over it’s short-term memory.

Jennifer Riggins and The New Stack crew have a good booklet out on Platform Engineering. I read over it and talked with Jennifer a couple times. I should have recorded those calls to munge into some articles, but, whatever.

You should check it out, I think it’s a good go at trying to nail down exactly what that term means. This month, at least :)

Wastebook

• The best way to make money is to not to loose money.

• Make It Bigger part 01 and part 02 - great stuff here, especially on focus (something I willfully ignore all the time) and building comfort with presenting. Also, Russell Davies has very unique, but subtle style to everything he does. I don’t know him, but I wonder how different his front-stage character is from his back-stage character. Is his vibe how he is in “real life”? He runs a conference called “Interesting” which I think is a word that summarizes his approach. Also, like me, he used to be very active and “big” in the 2000’s on the web and is, sort of, quiet now. How is that working out? I need a club of people like that where we meet and weekly answer the question “what am I doing now? should I be satisfied?” More broadly: the UK digital people are always, well, interesting.

• “it’s not not about the technology” Here.

• “Well, I don’t know how many of you have built Kubernetes-based apps. But one of the key pieces of feedback that we get is that it’s powerful. But it can be a little inscrutable for folks who haven’t grown up with a distributed systems background. The initial experience, that ‘wall of yaml,’ as we like to say, when you configure your first application can be a little bit daunting. And, I’m sorry about that. We never really intended folks to interact directly with that subsystem. It’s more or less developed a life of its own over time.” Craig McLuckie, SpringOne 2021.

• “It is a ghost of a word, with no meaning except that fact that it almost had a definition – now the definition cites it’s lack of existence.” Here.

• “Small enough to care.” Seen on the back of a semi-truck trailer on the way to Schiphol.

• If you tell me the procurement process for an organization I can tell you exactly what kind corporate culture it has. Enterprise Bloodwork.

• Privileged sad sack.

Relevant to your interests

• Longplay 2.0: An Album-Oriented Apple Music Player with Loads of New Features - Seems cool.

• Paper Calendars, quarterly ones - Print out A4 sized calendar pages, weekly ones, put until books

• Survey finds relatively few Americans actually use (or fear) ChatGPT - 30% to 40% (for young adult males) seem like a lot too me. // “Ongoing polling by Pew Research shows that although ChatGPT is gaining mindshare, only about 18% of Americans have ever actually used it. Of course that changes by demographic: Men, those 18–29 and the college educated are more likely to have used the system, though even among those groups it’s 30–40%.”

• The Ultimate Guide to Developer Counter-Productivity - Focus on outcomes, not activities…mostly. So long as the activities are good and lead to outcomes. // Yes, and…many tasks may not seem directly tied to value, but they are. If you don’t measure where a shelf goes on a wall, when you hang it, there’s a good chance it will look shit. If you don’t take your compliance training in a bank, the bank could be operating illegally. If you don’t take time to meet with other teams (people), when it comes to needing to work with them, it will be harder. Also, if the outcomes you’re unproductive developers are getting are greats, who cares about LoC.

• Man, Myth & Magic The Illustrated Encyclopedia Of Mythology 21 Volumes In 1 - This is the kind of stuff I’d pursue and spend hours looking through in the 80’s.

• Zodiac posters by Bruce Krefting, 1969

• Mindsets and Tactics for New Leaders of Software Teams - Seems like good stuff.

• Dark Mode: How Users Think About It and Issues to Avoid - A lot more on dark mode than you thought was possible to write about // “Aesthetic appeal and improved accessibility are the strongest arguments for supporting dark mode.”

• Here’s Why Automaticity Is Real Actually - How hype goes bad: “Suppose there is a good idea. People will be attracted to it. It will gather momentum. Eventually it will have made all the true claims it can make, but it won’t have used up its hype. Its momentum will carry it forward into making false claims and doing bad things.”

• Terraform vs. GitOps vs. System Initiative

Upcoming

Talks I’ll be giving, places I’ll be, things I’ll be doing, etc.

Sep 6th to 7th DevOpsDays Des Moines, speaking. Sep 13th, stackconf, Berlin. Sep 14th to 15th SREday, London, speaking (get 50% of registration with the code 50-SRE-DAY) Sep 18th to 19th SHIFT in Zadar, speaking. Oct 3rd Enterprise DevOps Techron, Utrecht, speaking. Nov 6th to 9th VMware Explore in Barcelona, speaking.

Logoff

I’m in Des Moines for DevOpsDays later this week, speaking on Friday. This is the view of downtown from my hotel room. It reminds me of Austin before all the new builds started coming in in the 2000s. They are BIG into agriculture ads in the airport. I don’t get that kind of regional focus on airport ads in Europe, so it’s always fun to be reminded of it.

Two airport thoughts:

1. In Europe, most airports I go to are in big cities, and big hubs. Those airports have a fancier feel. The staff in those airports also seem a lot cheerier than the staff in US airports. US airports have a very hostile feel to them from the sometimes ticketing, TSA, general staff, definitely customs and boarder patrol (they were nice this time! the only thing to Dutch ever do is make fun of me [rightly so] for not knowing Dutch after five years), etc. That said, American travelers are also more uppity - entitled even? I very rarely see people traveling in Europe getting upset or angry with staff.

2. Ben Thompson aside in some podcast awhile back that one of the reasons that American airports are so crappy (you know, old carpet) is that because they’re old. And they’re old because America was one of the first places to have lots of airports. I have no way to verify that, but it’s a good “turns out.” The turns out being that at one point, they were awesome and leading edge. First movers’ stuff becomes shit. I’d have to check. Also, as with so many things like this (comparing American, uh, “infrastructure style” to the rest of the world), the other thing is: America is a lot bigger than you’d think. We have a lot of airports compared to say, Germany, definitely The Netherlands, I wonder if all of the EU? (Something like 5,000 in the US vs. 3,500 in the EU - based on some shitty-quick Google searching).

The point being: I think they’re both just different systems and thus, have different results that are hard to compare. It’s a variant of “the purpose of a system is what it does.”Something more like “the system does/is what you choose.” So (a) realize it’s a choice, and, (b) choose wisely.

I’ve been experimenting/playing a lot with ChatGPT as a Dungeon Master. I like to call this “ChatDM.” That is, I’m a player, and it DMs for me. It’s not, like, great, but it is good enough if you want to co-imagine with it. Overtime, it is a great world building co-author with you. The current problem with using ChatGPT as a Dungeon Master is that it doesn’t have a long enough memory. I haven’t gone and checked the actual token limits and stuff, but it very quickly forgets things that happened long ago and will “hallucinate” past events, NPCs, etc. It does a good job, though! But it’s not accurate.

For example, through several playing sessions and then some text tuning with the ChatDM, I developed the idea and write-up of The Lost Tribe of Elderwood. I did this through a combination of me just making the stuff up, but also I had a few game sessions where the author of the text (Rye Leafoot) traveled to meet with The Lost Tribe to study them. Actually, I did it through another NPC, Sam the Tinker. I had ChatGPT play Sam and I “DM’ed” this. It was great, because the simplistic naïveté of ChatGPT worked to ask all the “dumb questions” that you need to go through to explore and brain storm for world building. And, it was super-fun! I started playing Sam walking into Aural’s Reach, meeting some lumberjacks, getting to know Rye Leafoot (the one who writes the monographs about Elderwood), and then played out traveling to meet The Lost Tribe and talking with them.

That memory problem is a…problem, though. If you can imagine doing this for, like four or five hours over several days, the short-term memory of the ChatDM is a problem. Until this problem is solved, the real problem that it’s not creative/imaginative enough can’t even be addressed.

What I’ve been doing to get over that problem is figuring out how to pass context back and forth between ChatGPT sessions. (The other issue is that you only get 50 interactions with ChatGPT 4 per three hours, so you’ll have to wait or shift-down to ChatGPT 3.5 which is much worse for ChatDM’ing.)

There’s two “context” passings that I do. First, I’ve typed up my own very briefing prompts to give a new ChatGPT session. I might should ask ChatGPT to fix these up, but I haven’t! To fit into the memory constraints, I’ve written several chunks that go from world, to city, to character, and then some other ones I’ll use as needed.

So, when I open a ChatGPT session to play Dungeons & Dragons, I give it this prompt to convert it to a ChatDM:

I want you to play a D&D 5e dungeon master. I will give you some brief explanations (“DM reports”) of the campaign setting. As we progress I will give you more DM Reports as needed. You will be inventive and creative in creating NPC interactions, encounters, adventures. You will control all NPCs and monsters and interact with me. Always ask me (the player) what to do next and never take actions on my behalf. Award XP when appropriate. If you have doubts about what to do, ask me how to decide.

I’ve seen a lot more detailed prompts that get detailed about combat mechanics and stuff. Many of those might be ChatGPT 3.5, but ChatGPT 4 is pretty good with the above.

Combat is still weird - sometimes I’ve fight a goblin (that typically has 7 hit points) and I have to ask how many hit points it actually has after doing 40 points of damage. Also, the ChatDM isn’t good at using monster tactics, like the goblins' strategy of making ranged attacks with a bow and then using other goblins to make hit and run attacks with nimble escape.

Then, I give it this prompt for my campaign setting, Elderwood:

This is the campaign setting: Elderwood is a dense forest of oak and pine. North is a steep long mountain range, The Elves’ Wall. West is a steep cliff drop-off to coastal planes like California coast, called The Royal Coast, runs along Sea of Passage. South of Eldergrove same drop off and stone forest. East more forest. The River runs diagonal through Elderwood, curvy and wide. In middle on north of The River is Aurel’s Reach. City of 15,000. Primary industry is lumber, artisanal crafts, supporting, guilds, trade bound west and east along The River. Farms across river support Aurel’s Reach. Govt is council of 7 elected. City is: old city in core that is pristine and well planned, ancient elven and human made. City expands from there in chaotic, haphazard walls built over 800 years woven in with forest buildings in trees, free standing, Spaghetti bowl of streets. Open air markets. Parchment row: book sellers row. The Whispering Willow tavern right on north most wall in town built half/half in giant redwood tree. Marna human owner, Bardin dwarf barman, Maria gnome waitress. The city is lively mix of all D&D creatures and races, coming and going, most people chaotic good or good of some type. Part of forest that surrounds Aurel’s Reach is Eldergrove. Rest of Elderwood is a typical D&D forest with magical glades, unexplained ancient ruins, dark parts of forest. Lumber camps and creeks and small villages throughout. Use forest encounter tables.

Since The Whispering Willow Tavern is so key to the campaign and adventures I do, I detail it a lot.

Then, I tell it about the character I play, Rye Leafoot:

I am playing a 5th level druid (of the Circle of Land, forest) lightfoot halfling named Rye Leafoot. He is chaotic good, curious, and known by many about town and in the forest. With the sage background his passion is writing monographs about Elderwood and writing on topics like National Geographic magazine. He rarely passes on the opportunity to learn more, question people and creates, and document them in his notebooks for research. He is a gourmand and keeps the halfling meal schedule (Breakfast 7 am. 2nd Breakfast 9 am Elevenses 11 am Luncheon 1 pm. Afternoon Tea 3 pm. High Tea 6 p.m. Supper 9 p.m. Pudding & Cordials 11 p.m). He lodges at and is well known at The Whispering Willow tavern/inn, has his own reserved table in back corner nook that he uses as a study and personal library. He has established relationships with some city guards, merchants of fine food (Alden the wine merchant, Tormin halfling cobbler, Elera elven ranger of city guard, Thanlin the alchemist, more). He has a passive perception of 19, passive investigation of 17, passive insight of 14. Other than those, I will do all other checks and rolls for Rye and dictate all his actions.

I tuned this a little bit as I started using it. The hafling meal plan is one of the fun parts of role playing for me, so I wanted to make sure it was core to ChatDM’ing the character. And, you can see I fed it some NPC friends.

So far, I haven’t had success with ChatDM using things like passive perception in the background, but I keep feeding it those stats. When I know a DM should make those checks, I tend to prompt ChatDM to do so. (And, come on, people: how can you passive perception be higher than your active passive - I’ve read the explanations, but it seems silly - I like the idea of that being a floor).

These three prompts are pretty good.

Once feeding these prompts in, the ChatDM likes to start an ultra-cheesy adventure. It often starts with the stereotypical thing of starting the game in the tavern. And far, the adventure always starts with Elera coming into the tavern with some problem (bleeding trees, an ghost lumber camp, etc.) In those cases (and when it does something else I think is cheesy), I’ll just write: never mind that, let’s go back.

As an example of specific information that I don’t feed it initially, here’s an entry about the Elderwood lumberjacks:

The lumberjacks of Elderwood generally are harmonious with the forest. They understand that they must maintain balance with the forest and this is what the Elderwood is so expansive and vibrant. Not all lumberjacks behave kindly to the forest but most do. Mistakes are made as well. Luberjacks typically go on month long jobs into the forest in groups of three or 18. Some lumber camps are long established. They typically fell trees and take them to the river to float them down as lumber barges, returning to Aurel’s Reach on a barge when done. Luberjacks follow the trope of blue collar hard workers a small number follow druidic religious thinking, all are superstitious and some know about Sylvan creates and lore. Lumberjacks are of all humanoid races and there are some rare saytrs and other humanoid Sylvan lumberjacks.

When I had it DM a recent adventure, when I got to the lumberjack camp, I fed it this information and it did pretty well.

I haven’t done this yet, but I want to try re-feeding the ChatDM this info when it forgets. I’ll say something like, “as a reminder, here’s info about Rye,” etc.

In addition to the campaign setting, I wanted a way to keep the DM’s memory about small things going. To do that, I have the ChatDM write “DM reports” about what happened in an adventure. For example, Rye befriended a pack of dire wolves in a recent adventure and, you know, I’d certainly like to do more with that. ChatDM will forget this hours later. So, I wanted to be able to feed it a report to remind it. I asked to to “write a report about what just happened (the encounter with the dire wolves) in 300 words.” It outputting this:

DMReport: Encounter with The Moonclaw Dire Wolves Date: Early Autumn, exact date to be confirmed Location: Northwestern parts of Elderwood, en route to investigate lumber camp anomalies. Summary: Rye Leafoot, a 5th level Druid, encountered a pack of four dire wolves known as the Moonclaw while traveling with Elera, an Elven ranger. Utilizing the "Speak with Animals" spell, Rye engaged in diplomatic communication with the pack leader and was led to an injured young wolf. Actions Taken: - Offered dried pork sausage to the Moonclaw as a token of goodwill. - Healed the injured young wolf using "Cure Wounds" spell, restoring it to full health. Outcome: - Gained the respect and gratitude of the Moonclaw pack. - Was told the location of a dangerous trap near a large oak by a creek. - Granted the privilege to return and meet with the Moonclaw in the future. Notes: - The pack leader expressed interest in future meetings. - Made a mental note to disable the mentioned trap upon return.

I haven’t used this yet, but I will soon. My hope is that through a combination of my own role playing, using some as you know Bob’ing, and the report above I can more or less restore the ChatDM’s memory of Rye’s relationship with the dire wolves. I saw it use TV Tropes when I asked it to profile some NPCs awhile ago and I have in the back of my head to start using those as references for characters. It seems to know them well, so that would be a great way to get around this memory problem, e.g., “Meribella is a ‘Jerk with a Heart of Gold’ dwarf.”

One mechanic I’ve used for this also is to have the ChatDM write an entry for Rye’s notebooks, or his monograph. To do this, I’ve had it analyize Rye’s writing style and general characrer and will feed it that as “in the style of.” You can see the effect of that in most of the Elderwood Guidebook entries, like The Whispering Willow one. The style matching is kind of arch, but good enough, especially for me to go in and rework it with minimal effort, like The Lost Tribe entry.

I’ll see how this goes.

Oh, one more thing. To add to something I wrote about, another theory I have is that after doing an adventure, you can ask the ChatDM to write an outline for a module out of what happened. It’s short memory will be a problem here, but what I’m hoping for is that I can create an adventure by playing it out and then ask ChatDM to spit out the outline. Then I can use that to start fleshing the adventure out more.