> I don't personally think there is magic in building a Docker container. Call me old-fashioned.

This seems like a false dichotomy. You don't have to do this. It is still possible to build magical things. But agents aren't it, I don't think.

It is honestly extremely depressing to read this coming from a founder of Relic. Relic built magic. Dawn of War and Company of Heroes formed an important part of my teenage years. I formed connections, spent thousands of hours enjoying them together with other people, and pushed myself hard to become one of the top 100 players on the CoH leaderboards. Those competitive multiplayer games taught me everything there was to know about self-improvement, and formed the basis of my growth as an individual - learning that if I put my mind to it, I could be among the best at something, informed my worldview and led me to a life of perpetually pushing myself to further self-improvement, and from there I learned to code, draw, and play music. All of that while being part of amazing communities where I formed friendships that lasted decades.

All of this to say, Relic was magic. The work Relic did profoundly impacted my life. I wonder if you really believe your current role, "building trust infrastructure for AI agents", is actually magic? That it's going to profoundly impact the lives of thousands or millions?

I'm sorry for the jumbled nature of this post. I am on my phone, so I can't organize my thoughts as well as I would like. I am grateful to you for founding Relic, and this post probably comes off stupidly combative and ungrateful. But I would simply like to pose to you, to have a long think if what you're doing now is really where the magic is.

Edit: On further consideration, it's not clear the newly-created account I'm responding to is actually Alex Garden. The idea of potentially relating this personal anecdote to an impersonator is rather embarrassing, but I will nonetheless leave this up in the hope that if there are people who built magical things reading this, regardless of whether they're Alex Garden or someone else, that it might just inspire them to introspection about what building magic means, about the impact software can have on people's lives even if you don't see it, and whether this "agent" stuff is really it.

>The idea of potentially relating this personal anecdote to an impersonator is rather embarrassing

Good news! You've also related it to the roughly ~3-10M monthly HN readers who are not (potentially) impersonating the founder of a beloved game studio.

Also: I think you're probably safe. I'm sure someone at some point has come to HN to LARP as some prominent person in tech that they don't happen, at that specific moment, to actually be... but I can't really think of it happening before, nor would I expect it to take the form of a particularly thoughtful comment if a troll did that. Though with AI these days, who knows? I might myself just be one of a swarm of clawd/molt/claw things. In which case I'd be the last to even know it.

Oh-- as for being depressed about their docker/wiring things up sentiment. Try not to be, and instead, consider: Is it a surprise that someone who founded such a place as relic was occasionally-- even often-- frustrated at the things they had to clear away to build the thing they actually wanted to build? People who want to build amazing experiences may not love having to clear clutter that gets in their way. Other people want to build the tools that clear clutter, or other things that keep the whole system going. Those are beautiful too.

Oh man, modding Company of Heroes was one of the things that got me into programming. I look back fondly on those memories.

If we've arrived at the point where bots are impersonating me (instead of the billions of other choices), I'm probably at peak Alex. I'll light a candle. So... easy to disambiguate this one.

I got the idea for Homeworld one night when I was about 21. At the time, I was working at EA as a programmer on Triple Play 98 (building FE gfx - not glamorous). In an RTS-ironic twist of fate, my boss and mentor at the time was Chris Taylor - go figure.

Friends of mine had their own game company and had boxed themselves into a technical corner they couldn't get out of, so I agreed to write a bunch of sprite conversion code for them after hours. That night, we were all working in a room, talking about the reasons X-Wing vs. Tie Fighter didn't work on a 2D screen (hold up and left till you turn inside and shoot) and how Battlestar Galactica didn't get the cred it deserved, and BOOM - in my mind I saw ships in 3D with trails behind them. Inside a crystal sphere like Ptolomy's theory of the universe (man inside - god outside), and I saw that the surface of a sphere is 2D, so you could orbit OUTSIDE with a mouse... it looked like spaghetti floating in zero g... that's why Homeworld's working title was "Spaghetti Ball" for months.

Fortunately for me, in this ambiguous thread, I can give you all the proof of life you want. Try me.

Now... is transparent and trustworthy casting spells? Yeah... it is, but not by itself. It's a primitive - a building block. My personal projects (that I do think are magical) kept running into the same problems. Effectively, "how do I give up the keys if I don't really know what the driver is going to do?" I tried coming at this problem 10 different ways, and they all ended up in the same place.

So I decided to go back to the basics - the putpixel(x,y) of agentic workflows, and that led me to transparency and trust. And now, the things I'm building feel magical AND sustainable. Fun. Fast... and getting faster. I love that.

At Relic, our internal design philosophy was "One Revolutionary and Multiple Evolutionary". The idea was that if you tried to do more than one mind-blowing new thing at a time, the game started feeling like work. You can see this in the evolution of design from Homeworld to DoW to CoH (and in IC too, but let's face it, that game had issues <-- my fault).

Now... on the topic of "Is agentic coding better or worse", I feel like that's asking "is coding in assembler better or worse". The answer (at least used to be) "it depends"... You're on a continuum, deciding between traditional engineering (tightly controlled and 100% knowable) and multi-agentic coding (1,000x more productive but taking a lot for granted). I've found meaning here by accepting that full-power multi-agentic harnesses (I rolled my own - it's fucking awesome) turn software engineering into Socratic debate and philosophy.

I don't think it's better. It's just different, and it lets you do different things.