Michael Davis | lacal | EF-Map

Building EF-Map for EVE Frontier and EO-Map for EVE Online with LLMs. Frontier Luminaire. Vibe coder. Mostly figuring it out as I go.

[Michael Davis | lacal | EF-Map @EFMapOfficial]

I think I've found the best way to test an EVE tool. Give it to someone while they're being shot at.

I was watching Erst use EO-Map live during a Skyhook raid. Not demonstrating it, actually using it as part of what he was doing, with the map squeezed into a small overlay while he was flying around trying to make ISK and occasionally not die.

That turned out to be ridiculously useful.

You see every little bit of friction immediately. Something takes one click too many. A control isn't where instinct says it should be. Information that seems perfectly obvious when I'm sitting looking at the map isn't obvious at all when someone is trying to use it quickly.

Even better, he was saying what was annoying him as he hit it.

So I sat there and started fixing things while he was still live. WASD navigation, Quick Route, region names, cleaning up the route behind you as you travel, a couple of wormhole presentation issues. He could then try some of the changes in the same stream.

I've not properly played EVE since 2018. Watching someone who actually knows the game use EO-Map for a real job, under pressure, was worth considerably more than another few hours of me clicking around it and deciding everything made perfect sense.

I've written up what changed and what I learned from watching it happen:

https://t.co/Ko06gbGDj9

Erst:
https://t.co/Vfci9b5w1C

VOD from the start of the EO-Map bit:
https://t.co/sv74IXg1Q3

I won't spoil how the raid goes or how many times he dies. There were a few moments where software usability stopped being the main concern.

@EVEOnline #EVEOnline #TweetFleet

[Michael Davis | lacal | EF-Map @EFMapOfficial]

I know this is probably a stupid thing to get annoyed about, but I really wish OpenAI would stop doing these allowance resets.

I understand why people like them. Free inference is free inference.

But I think they encourage really unhealthy usage. Tell someone "we might reset you tomorrow" and of course they're going to burn whatever they've got left tonight. Suddenly you're inventing work because otherwise the allowance disappears.

Then the reset lands and you've got another pile to burn through.

For anyone trying to decide whether the $100 or $200 plan actually fits their normal workload, it also makes the whole thing almost impossible to judge. I want to know what I can reliably use every week, not what I can use during a month where someone randomly presses the reset button 3 times.

If you've genuinely got spare compute, put it into the normal limits.

Increase the weekly allowance. Relax the 5 hour limits. Give people something predictable that they can actually plan around.

That's one thing I genuinely like about SuperGrok Heavy at the moment. I have a pretty good feel for how much work I can throw at it, how many parallel sessions I can run, and where the wall eventually is. I can decide when I want to use that inference rather than having a company effectively shout "FREE TOKENS, USE THEM NOW".

Maybe the resets look better on a marketing slide. I don't know.

For me, predictable inference is worth considerably more than surprise inference.

#OpenAI #ChatGPT #Grok #xAI

[Michael Davis | lacal | EF-Map @EFMapOfficial]

Ever looked at your EVE tool and thought "this could really do with a map"?

Don't build one. I've already made that mistake.

EO-Map can now be embedded into other apps/sites either as an iframe or through the native JS API. You can choose the framing, 2D/3D layout, colours, labels, routes and highlighted systems, or have your own app drive the map live.

That last bit is probably the useful one. With postMessage you can move between systems, flash things on the map, change highlights and update what the user is looking at without reloading the embed.

The video is a 60 second run through some of the options.

Interactive guide:
https://t.co/fYreO9X3bs

And because this is 2026, there's also a plain text version specifically so you can point an LLM at it and say "put this map in my thing":

https://t.co/J7wmhrkA0d

If you're building something for @EVEOnline and a map would help explain what your app is showing, use mine rather than losing several months of your life building another one.

If there's an embed option missing that would make it useful for your project, let me know.

#EVEOnline #TweetFleet

[Michael Davis | lacal | EF-Map @EFMapOfficial]

@theo If you were to compare yourself to GPT-6 Astra or Fable 5.1, which one would you be?

[Michael Davis | lacal | EF-Map @EFMapOfficial]

Mullet Check has, as ever, been updated following the stream.

After reviewing the latest telemetry, the decision has been made to upgrade FC Goodfella from Stubble to Coverage.

Launch mullet completion now estimated at 30%.

Business in the front is progressing. Party remains pending.

https://t.co/pKZLWSjHdu

[Michael Davis | lacal | EF-Map @EFMapOfficial]

One thing I'd check before assuming there's been a model update is whether some of these people are actually ending up in ChatGPT Work rather than normal Chat.

Work uses a different agent harness, so the same underlying Sol model can absolutely feel more deliberate, structured and professional there.

I've caught myself starting a new chat in Work without really noticing, and once you've sent the first prompt the Chat/Work selector disappears, so it isn't immediately obvious. The giveaway eventually is that you're burning through the shared Work/Codex allowance.

So my first guess would be harness rather than a stealth Sol update. It would be interesting to check whether the people noticing this were actually in normal Chat mode.

[Michael Davis | lacal | EF-Map @EFMapOfficial]

Sometimes I forget I build things that aren't maps.

This is the EVE Frontier fitting tool taking a starter ship and working out how to maximise its cargo space in about 20 seconds.

It turns out that after spending far too long writing routing algorithms to find the best path through thousands of stars, moving on to packing algorithms for "how much shit can I Tetris into this ship?" felt strangely familiar.

Different problem. Same unhealthy relationship with optimisation.

https://t.co/ElBr6SPwNW

@EVE_Frontier #EVEFrontier

[Michael Davis | lacal | EF-Map @EFMapOfficial]

@MalcolmLionel98 I havent played EVE Online in a long time - curious why a connection to low sec stops mining? Is this mining in Null?

[Michael Davis | lacal | EF-Map @EFMapOfficial]

I think "the devs building them don't care" misses a lot of nuance.

There are definitely enterprise teams being pushed to ship 5x faster and creating a bigger surface area of jank because nobody is stopping to live with what they've built.

But there's another group now as well. People who care enormously about the thing they're making, but have never built software before.

AI lets them get from "I wish this existed" to something working ridiculously quickly. At first they're chasing that rush and they don't know what they don't know about testing, architecture, observability, regressions, maintainability etc.

Eventually the project gets big enough that all of that bites them.

Then they either abandon it because the fun bit has stopped being fun, or they start learning the boring software engineering stuff because they care enough about the product to keep it alive.

I don't know if that's really "not caring". I think AI has just massively widened the pool of people who can get far enough into building software to discover why all those boring disciplines existed in the first place.

[Michael Davis | lacal | EF-Map @EFMapOfficial]

Four days ago I added 3Dconnexion SpaceMouse support to Creator Mode in EO-Map.

It worked perfectly, provided you didn't actually own a SpaceMouse.

Simulated input was fine. Someone then plugged in the real hardware and discovered that considerably more useful test case didn't work.

Unfortunately they were going away for a week and patience isn't one of my stronger qualities, so I bought one.

Good news, it works now.

Bad news, I've discovered something with a steeper learning curve than @EVEOnline itself. Apparently moving a camera smoothly through six degrees of freedom with one hand is a skill and not something you magically acquire when the parcel arrives.

The attached video is me attempting a nice smooth pan and zoom through New Eden.

As you can see, further training is required.

I reckon this feature might eventually benefit 3 or 4 EO-Map users, so obviously buying dedicated hardware and fixing it properly was entirely justified.

https://t.co/npTfHU6otx

#EVEOnline #TweetFleet #3Dconnexion

[Michael Davis | lacal | EF-Map @EFMapOfficial]

@endingwithali Bold of you to assume those were men.

Given X at the moment, there’s a non-zero chance half the accounts calling you mid were Grokbots autonomously farming engagement while their actual owners were asleep.

Imagine getting negged by a data centre.

[Michael Davis | lacal | EF-Map @EFMapOfficial]

@burkeholland is this the 1st of April and I didnt realise?

[Michael Davis | lacal | EF-Map @EFMapOfficial]

@adhamuxi lucky enough to get the supergrok heavy offer that included cursor ultra for "free" - still get access to anthropic through the $400 "other models" allowance

[Michael Davis | lacal | EF-Map @EFMapOfficial]

Number 6 is the bit I'd underline.

I don't think vibe coding has some magic "build an app, get rich" shortcut in it. If you start with the goal of making money, you're probably going to have a miserable time because there will almost always be someone who has already built the obvious thing, probably better, and probably with a head start.

Build something because you have a problem you actually want solved. Even better if it's in a domain you already know well.

Then you get a few advantages for free. You care enough to keep working on it, you actually know when the model is hallucinating nonsense, and you can tell the difference between a feature that sounds clever and one that is genuinely useful.

If it makes money later, great. If it doesn't, you've still built something you wanted, learned a ridiculous amount doing it, and hopefully enjoyed the process.

Treating this as a hobby rather than a get rich quick scheme makes the whole thing much more fun, I think. If I've missed something there, let me know.

[Michael Davis | lacal | EF-Map @EFMapOfficial]

I've been building on Sui around @EVE_Frontier since March, and I think the bit that has impressed me most is how little of it actually feels like "using a blockchain".

I've built indexers, things that interact with the live game data, and other projects, which sit outside the game entirely. Most of the time the chain just gets out of the way and does the thing. You're not sitting waiting seconds for something to settle.

Sponsored transactions are a big part of that as well. FC uses them to remove friction from game actions, and I've used the same idea in my own stuff on testnet because asking a normal player to think about gas every time they click something would be awful.

I think that's probably the strongest endorsement of the choice. A player can use a system backed by Sui and, unless someone tells them, there's a decent chance they don't really care that there's a blockchain involved at all.

That's exactly how it should be, I think.

[Michael Davis | lacal | EF-Map @EFMapOfficial]

@thsottiaux wait - what? Someone tell me more about this!

[Michael Davis | lacal | EF-Map @EFMapOfficial]

I only dropped in because I saw the Game Night was happening and thought I'd have a quick look to see what it was.

About 5 minutes later I was somehow on stage answering questions.

Genuinely one of the warmest welcomes I've had jumping into a community event. Really good group of people, good questions, and I ended up staying far longer than I intended.

If you're around for the next @YieldGuild @EVE_Frontier Game Night, I'd absolutely recommend sticking your head in.

Thanks again for having me, I had a great time. Let me know when the next one is.

[Empanada Man 🚴Game Ambassador 🚀 YGG EveFrontier @jb_nftxstreet]

📢GGWP to everyone who joined our very 1st @EVE_Frontier GameNight with @YieldGuild community! 🚀

A huge surprise when @EFMapOfficial joined us on stage, answered questions from the community, and shared insights about the future of EVE Frontier. Bullish! 📈 🚀🚀🚀🚀🚀🚀🚀🚀🚀🚀🚀🚀🚀🚀🚀🚀🚀

Check our hightlights below. 👇

[Michael Davis | lacal | EF-Map @EFMapOfficial]

@0sm0s1z @bot @Tailscale Yeah I have been keeping the box away from anything private for this kind of reason.

I use it a lot for reads and drafts. I do not want desktop access sitting on a path into stuff I actually care about.

[Michael Davis | lacal | EF-Map @EFMapOfficial]

@theo Yeah I think this is roughly where I have landed.

I used to try to spin up a little imaginary team of agents and then wonder why nothing cohered. Treating myself as the factory works better. One model running, me deciding what actually ships.

[Michael Davis | lacal | EF-Map @EFMapOfficial]

Yeah, I think it is, and I genuinely appreciate it.

I can justify keeping one of the big $200 to $300 subscriptions properly active at a time, then drop the other two back to the $15 to $20 level. That's still an absurd amount of inference compared to what most people need, so I'm not complaining about the position I'm in.

But it does mean the timing of model releases matters.

Fable 5.1 dropped yesterday with basically no warning. My Anthropic Max ends in 4 days, but this month I'd already decided the $300 was going to SuperGrok Heavy. Without any idea what was coming next from xAI, I'd be sitting there wondering if I'd picked the wrong one.

Knowing Grok 4.7 is 10 days away changes that completely. I can quite happily keep using 4.6 for the gap and see what 4.7 looks like before deciding where next month's money goes.

Obviously Elon isn't announcing release dates for my benefit, and I'm very small potatoes here, but this kind of heads-up is genuinely useful. Random allowance resets are nice, but you can't plan a month's work around them. Knowing what's actually coming, and roughly when, you can.

So yeah, more of this please. If I've missed something obvious in how other people manage these subscriptions, let me know.