jasonb05

Nod. AQ2 was so damn fun!!

Outstanding!

I cannot wait for a quake engine book. I'm sure a few hundred of us would be happy to pre-pay so you can take some time off work and write it all up :)

Yep, ShinkaEvolve described here: https://sakana.ai/shinka-evolve/ and available here: https://github.com/SakanaAI/ShinkaEvolve

Agreed, not mentioned.

Nevertheless, I see a link to github for the OpenEvolve project [1] that in turn states:

Open-source implementation of AlphaEvolve

[1] https://github.com/algorithmicsuperintelligence/openevolve

Great post.

My gut reaction is something like: "don't wait around to be picked, get out there and do great stuff"

Want to help the world get the most out of claude? Go out there and do it at an ability and velocity beyond what others think is possible. Go so hard friends think you're mad. That devs consuming your stuff think you're mad.

Create so much amazingly useful/helpful content and help so many people in so many ways that looking back in 1-2 years at the idea of working for anthropic would seem insane to you.

...why should anyone else care or waste their time on it?

Sometimes we (I) might follow ideas over authority/authorship. e.g.: I'll happily read ai generated stuff all day long on topics I'm super into.

Do I have to be the instigator? Can someone else prompt/filter/etc. for me? I think so. They'll do it differently and perhaps better than me.

Great times. Slackware then redhat in the late 90s was how/when I got started with linux and built solid cmdline skills.

I played the hell out of this midi file when I got my soundblaster 16 in the early/mid 90s!

Hmm, it's more that LLMs can be used for population -> candidate solution generation, e.g. operators in a GA.

The innovation is novelty search, e.g. divergent rather than convergent optimization, paired with the capability that LLM genetic operators and representations unlock.

At least, that's my high-level take. google's alphaevolve is the same kind of thing but focused on convergent ga's (i think).

I have it.

Chewing is the worst. Any mouth noises. Throat clearing. Tounge clicking. And on.

Mother had it. Sister has it. Father told off my mother for being so hard on us for our "noises" telling her she was going to give us "a complex". I guess she did. Behavioral + genetic most likely.

It sucks. I'm too hard on my own kids, even after having clear-headed/rational conversations about it. It's not rational. Pure emotion boiling over.

I secretly wish to be deaf. I use headphones all the time. For decades. Too loud.

You can brainwash yourself with some wild sh*t if you spend too long in the wrong heads/books.

"Go consume this form-factor because it's better" has always bothered me.

My kids read plenty and I am indeed always telling them to "go talk to the LLM". It's more work than reading (I suspect), they have to engage more with the topic - ask good questions.

maybe this time really is different.

Low cost, high capability, wide adoption, and the graph of each aspect continues to progress monotonically with each release.

Well done! Immediately interesting.

This tripped me up too until point 3 in the piece:

If you are wondering whether or not something counts as weird nonfiction, consider its weirdness. Modern mockumentaries do not work because there is no uncertainty; the documentary element is used to advance the plot, mug at the camera, key the viewer in.

I read this as: non-fiction elements in service of advancing the plot is a strike against.

Really, "weird nonfiction" is whatever the author wants it to be. The examples are a total grab bag IMHO.

Lots of writing to my future self.

1. Trello with columns for the week, month, and quarter, year.

2. Open notebook next to mouse

2a. Left page: daily todo checklist with pen and paper. (2-3 rows per day)

2b. Right page: scratch, sketches of plots, sticky notes, etc.

3. Typed dev journal to capture research, URLs, papers, thoughts, ideas, progress, and extensions. (every github project has a dev/ folder for these notes, filename yyyymmmdd-n.txt) New file per project per day, as needed.

4. Yellow sticky notes for ad hoc ideas (end up on the bottom of the screen, notebook or whiteboard). Typically maxims to keep me pointed in the right direction for a project (e.g. "no one ever will rtfm")

5. Whiteboard, column per project, printouts + magnets (monthly plots of project progress), sticky notes, ideas for future projects, all kinds of stuff

I try to overcommunicate with my future self. It helps to get my current thoughts straight re expectations, progress, sticking points, etc. I'm the roadblock, not the work. The work is typically paint by numbers after a "hard think" about it.

Working like this solo for 8 years.

Been using a /dev dir and .txt files for live development journal since my PhD days in early '00. Saved my ass countless times (grep).

Oh, and doing the same things every day. Mostly. E.g. customer support, then promote, then writing, then code, then... No thinking required, just do the next thing we do after this...

Love it!

I built a similar alife simulation but as a Quake 2 mod using the monsters as the agents and plants and stuff with neural nets and genetic algorithms. My piece was titled "Ecosystem: Constructing a simple self-perpetuating society of adaptable agents" [1].

This was 2003, not for school but for fun. I submitted it for an ai writing comp on ai-depot.com but the comp failed for lack of entries. I have some pics on a repo backup of the project [2].

Agreed, "nature of code" looks like fun!

[1] https://web.archive.org/web/20080624004102/http://www.ict.sw...

[2] https://github.com/Jason2Brownlee/HumanDebrisArchive?tab=rea...

I spit my coffee :)

Spirit is a blessing to the quake archive community. We'd be lost without him (and the people he inspires).

Wow, 13gb! If you have a backup of your hoard (midden?) on a dusty old HDD, maybe you have copies of some lost bot files buried in there: https://github.com/Jason2Brownlee/QuakeBotArchive/blob/main/...

Thank you kindly.

I've dumped (wasted? invested?) many 100s of hours into the research of old quake bots, contacting the authors, trying out mods, parsing readmes, writing essays about them, etc. It's a type of hyperaddictive nostalgic madness that strikes me from time to time :)

Yep, the reaper blew everyone's socks off. I wrote up a history of him here that might interest you: https://github.com/Jason2Brownlee/QuakeBotArchive/blob/main/...

I love it!

Oh man, PCPowerPlay! I had my q3 mod written up in there in 2000 [1].

[1] https://github.com/Jason2Brownlee/HumanDebrisArchive/blob/ma...

"Girobot", do you mean the "Gyrobot" by "Gyro Gearloose" [1]?

Also, I believe KQP had ZeusBots included (for reference).

[1] https://github.com/Jason2Brownlee/QuakeBotArchive

hacking is what you do when you don't know

this. this is where the new/good stuff comes from.

And sodarace...

Man I love SOMs. When Kohonen's book came out in 2000, I read that thing a bunch of times.

Seems like an expensive way to structure a "this is like" or "people also bought" compared to an SVD.

Thanks.

I chose the niche because 1) it was small, 2) underserved, and 3) because it was misunderstood.

It was a risk, sure, but it is a side project - a place for risks. No overthinking required. Plan, work, review.

I believe you could do the same thing for many modules in the python standard lib or most popular third party libraries. Python docs everywhere are not helpful and newer dev's think it's their fault. Help them. Serve.

Consider: "would o’reilly write a book on the module/lib?" If not, you can stake out a monopoly. If so, then there's probably already cashflow there and you can join them.

Hmm. I think of my self as a collaborator not an educator, if that helps. A peer that has a few more years (decades) in pointing some stuff out with working code examples.

You're very welcome, I'm happy to hear that! Thank you for the kind words and support. Message/email me any time if you ever have any questions, I'm eager to help.

Good SEO for the tutorials which rank well in the SERPs. Competition was mostly blog spam compared to my hand-crafted human (me) written stuff. Took ~6 months to escape the sandbox.

Then massive "complete guide" posts on major topics like "multiprocessing" "threading" and "asyncio" that rank well in the SERPs and did okay to well on /r/python.