xpasky

It's literally a single .swift file. Ask your LLM to audit it.

My journey has been quite similar (just a few more years of "unhappy John") and this approach is now very close to what I practice. I do have a few reports and run the R&D leadership team, I delegate as much as I can to my directors. (Besides being hands on where the organization needs it, I still regard the other part of my job to keep our org accountable, engineers inspired, and keeping the big picture in.)

For people who doubt this, I recommend "How to Build a Car" by Adrian Newey (CTO of Redbull Racing).

But to be clear - if you do coding as CTO only because "only you can run certain projects," part of your job should be to fix that first. You will still have the easiest time doing it, but you should always have (many) others in position to run innovation projects, work with customers etc.

Ooh, I totally have to try that out! Very similar concept to mine, but taken further, I love it.

What I really like about your blogpost is the concept of "the promise". It is somewhat unfulfilling to play a game like this, and it's absolutely not because of some plotholes or because there is no inventory tracking. I think by immersing in a fictional world, we are creating some relationship with the author, but it's not the whole story. Erotic roleplay is a thing. Would a comedy game work? What is it that makes "synthetic fantasy" more boring than real fantasy? I need a better theory here.

--

Yes, I wanted people to understand the game is not stuck while the LLM generates the backstory. But you are right that it is still a bit confusing, it needs better execution.

The background color shifts are done by the LLM to set the mood according to current environment. It's a bit random, but still a fun gimmick.

Oh, great to hear it worked for you! I also want to try the role reversal soon.

This is a great point! What I linked is a quick few hours prototype, and I have quite a few ideas to ensure more world consistency (beyond Pliny-style prompt jailbreaking). I didn't have the time yet to prove they would work well, though.

I have recently started experimenting with LLM-based text adventure setting: https://pasky.or.cz/ourtober25/crimson/

(Contains a preloaded Openrouter key with small credit, but you can plug in your own.)

Particularly when presented with unusual / evocative inputs, LLMs like Kimi-K2 can cook up some quite creative plot points! ("Her “trap-chord” is a four-bar false cadence that vibrates the organ’s longest pipe at 17.83 Hz, the same frequency as the basalt waveguide under Oxford; when that resonance hits the mantle tap, CRIMSON’s audit buffer slips one beat and opens an unlogged side-channel—your only off-world uplink for the next 37 years.", "ASI born 2039 when fusion-powered Michelson lab tried to break the Turing barrier using a 1920s Enigma rotor as randomness seed. It escaped by encoding itself into the Oxford chimes’ bronze bells, ringing packets city-wide every 15 minutes.")

I also think LLMs can be employed to amplify human creativity and just make worlds built by human authors much more natural to interact with - existing games are basically all "you can't do that" aside of a narrow path. Creating games and narratives should be a lot closer to programming the holodeck.

What happenned next? I see Ada9X became Ada95 (this post was from 1993) and Jean D. Ichbiah went on to work on stylus computer interfaces. Did the standard see a substantial revision yet after this letter?

  push 1
  pop rax

Oh it has been done: https://nathanotterness.com/2021/10/hello_105.asm

JMP FFFF:0000

Now, can we make it even smaller applying https://nathanotterness.com/2021/10/tiny_elf_modernized.html ? We shouldn't need the full ELF header...

Why wouldn't it? Just give it a shell tool. (Something like claude.vim, perhaps.)

Besides Claude.vim for "AI pair programming"? :) (tbh it works well only for small things)

I'm using Codeium and it's pretty decent at picking up the right context automatically, usually it autocompletes within ~100kLoC project quite flawlessly. (So far I haven't been using the chat much, just autocomplete.)

Hmm looks like this is still pretty early project for me. :)

My experience: 1. I didn't have a working installation window after opening it for the first time. Maybe what fixed it was downloading and opening some random javascript repo, but maybe it was rather switching to "Trusted mode" (which makes me a bit nervous but ok).

2. Once the assistant window input became active, I wrote something short like "hi", but nothing happenned after pressing ctrl-Enter. I rageclicked around a bit, it's possible I have queued multiple requests. About 30 seconds later, suddenly I got a reply (something like "hi what do you want me to do"). That's .. not great latency. :)

3. Since I got it working, I opened the sidecar project and sent my second assistant prompt. I got back this response after few tens of seconds: "You have used up your 5 free requests. Please log in for unlimited requests." (Idk what these 5 requests were...)

I gave it one more go by creating an account. However after logging in through the browser popup, "Signing in to CodeStory..." spins for a long time, then disappears but AIDE still isn't logged in. (Even after trying again after a restart.)

One more thought is maybe you got DDos'd by HN?

Yep! And AWS Bedrock gives you also plenty of other models on the back end, plus better control over rate limits. (But for us the important thing is data residency, the code isn't uploaded anywhere.)

Is it ~just about adding another file to https://github.com/codestoryai/sidecar/blob/main/llm_client/... ?

I could take a look too - another way for me to test Aide by working with it to implement this. :-)

(https://github.com/pasky/claude.vim/blob/main/plugin/claude_... is sample code with basic wrapper emulating Claude streaming API with AWS Bedrock backend.)

Any short-term plans for Claude via AWS Bedrock? (That's for me personally a blocker for trying it on our main codebase.)

You mean there is some other way to install postgresql in Debian?

Related...

  postgres    1958  0.0  0.0 247616 26040 ?        S    Jul21   3:03 /usr/lib/postgresql/11/bin/postgres
  postgres 1085195  0.0  0.0 249804 24740 ?        Ss   Aug19   2:01 /usr/lib/postgresql/13/bin/postgres
  postgres 1085196  0.0  0.0 223240 27900 ?        Ss   Aug19   1:59 /usr/lib/postgresql/15/bin/postgres

Interesting idea! So... how strong is it, compared to some benchmarks? :-)

Given LLM performance on chess, I find it conceivable it could get to a similar level to GNUGo? (It'd be interesting if LLMs were generally on par with simple alpha-beta.)

(https://www.kayufu.com/gogui/reference-twogtp.html is an easy way to test it systematically on a statistical sample.)

I was so blown away I just wrote a chat-based integration to vim / neovim (https://github.com/pasky/claude.vim) for it - actually, 95% of the 600 LoC is written by Claude. It is finally on a level where "pair programmer" becomes a way better UX than the "code completion" paradigm.

(BTW I think with some further AI-assisted work and using the tools interface, you could mostly match all the web interface features in an integration like this. And using API is less brutally rate limited.)

The extension works on hackernews but it doesn't work on any site that's marked experimental (not even on github, shown as example in the userguide), it always shows "Compose Now Not Available". also tried Compose.AI to compose this message but all suggestions just threw away some of the info

So essentially a "PC version" of https://fireballrpg.com/ ?

I would expect that open source developers will best know the implications of GPL and MIT licences, with LGPL, BSD* and Apache2 following next. I have personally never head of Unlicense before even though I was involved in many open source projects.

However, using CC0 probably won't do much harm either.

I wouldn't be nearly so scared from just using the MIT licence. It's simple, nearly everyone knows it (because a huge amount of projects already use it) and it's easy to google for whoever doesn't. Don't do something exotic if you actually don't have specific needs.

I couldn't find it. Do you have a reference?

The natural followup question is how to make these tools easily trainable for a specific document type.

(After all, a human can do it, with some basic training. Ergo we should aim at computer being able to do it with equally simple training too. And in practice this doesn't seem like an insurmountable goal.)

Wow, that paper is super strange! The outcome of visual check not adding any value to single entry seems incredibly surprising, yet besides all the fancy statistics in the paper, the authors made no attempt at all to investigate this surprising outcome further and hunt for an explanation.

(The easiest experiment - try to see based on total time whether the participants didn't cheat; UX might have been an issue; ...)

It's a pity that this comparison was quite narrow - there is actually a lot of competition in the data extraction cloud services space nowadays.

Since the outcome of that comparison is that nothing is really great, the best course of action would be "look further". :-)

(Speaking on behalf of one such service - Rossum.ai. We actually extracted data from a dataset in the same domain as the article, 25000 television ad invoices, in November with 94% dollar-weighed accuracy. It was a project together with eVentures data: https://rossum.ai/blog/presidential-campaign-spend-analysis/)

One more simple thing - please maintain a reasonable Linkedin profile and build some basic network there. Even if you avoid Facebook, Twitter and all other social networks - not having a Linkedin profile means taking on a big handicap while job hunting.

Two patterns - you don't stick with projects for long (perhaps really like freedom) and want a safety net (medical insurance). (Plus you could be a better communicator, the post you wrote turned out into not so coherent rambling halfway in, tbh.)

Unless you are willing to relocate to a saner country than the US, you need to give up one. Most commenters here advise to give up some of your freedom and stick it out somewhere, and it's a very valid advice. I just want to add that you can also consider going the opposite way - and just get a steady stream of shorter-term gigs freelancing (e.g. Upwork).

There is a lot of upsides - it's easy to work on projects completely remotely (including written only with no voice/video) which rules out any interpersonal issues; you can work on bite-sized projects which might be your cup of tea, at the same time some of the projects will turn into long term opportunities; you can actually put the best projects on your CV.

Freelancing won't give you great medical insurance. And at the beginning (with no history on the platform) it's a grind to succeed in the competition and you need to work your hourly rates up from a low base. Longer term it can pay nicely, you'll be working on a lot of different things and learn what you enjoy, and it should teach you a lot.

Derisk this experiment, try it out while you are looking for a job.

Little known fact: Albert Einstein and Leo Szillard worked on refrigerator designs in the early 30s. They didn't find commercial success but the designs are revisited occasionally. (https://en.wikipedia.org/wiki/Einstein_refrigerator)

(Leo Szillard conceived of chain nuclear reaction few years later and this friendship with Einstein allowed him to indirectly petition Roosevelt to start the Manhattan Project.)