e12e

Even if self hosted?

https://netboot.xyz/docs/selfhosting/

They already are Cyberdyne Systems IRL?

https://en.hyundai-wia.com/business/defense_business.asp

> it keeps your app and data in one place (I remember having some issue with this at the time, though I can't remember what the actual issue was)

CORS headers?

Correction - not "my" but "mu" (mail utils):

https://www.djcbsoftware.nl/code/mu/

Btw, anyone still reading/participating mailing lists? From MacOS? Have you found a reasonable client?

I'm not even going to try with Outlook.

Surprisingly good is a stretch. Barley adequate more like it.

Now that they've hidden mail access behind oauth (imap and SMTP, additionally SMTP behind global default off policy) and graph api behind oauth2 - it looks like they don't have to worry about real mail clients competing.

Actually fighting [f] to get mail in/out working with freescout right now - and having had learn more than I care to about o365 and PowerShell etc - I wonder how hard it would be to write a couple of stand alone tools to get fetch/send/sync mail working with o365 and local maildir - to get my/sup/any sane Mua to really work with o365/exchange/outlook.

Then there's calendar and teams to deal with..

[f] Thankfully our o365 reseller does most of the fighting - I'm happy to not have tenant-wide admin in AD/entra/whatever kerberized LDAP is called today.

The 30% assume gross, pre tax income - not net.

Ed: obviously the 50/30/20 rule sums to 100% so would need to be net - although the US is weirdly treating health insurance as a "premium service" that go into the 50% here, not into taxes like in a modern society.

Write one (system) to throw away...

How so? Let's say that over a year, a given section of code needs to be read and understood once a month. Taking some time to keep the code succinct and free of distraction will increase productivity all those occasions, as well as the rest of the lifetime of the system. Say the next decade.

How is that not efficient?

Great article. I'm not sure the author is correct - but I think something is happening to the adage:

> A sufficiently detailed specification is runnable code.

In a way I think LLMs will enable the dream of 4gl and "sufficiently smart compilers"[c].

LLMs aren't smart, but they are capable. Especially capable of translation and transformation.

I can certainly see them help move the abstraction horizon at which we work - so that rigid high level descriptions of the desired logic/process along with the process for quality testing - become the relevant curated artifacts - and the generated go/rust/java/python/etc code become incidental and mutable; subject to constant rewriting as part of the deployment of systems.

[c] You know, the ones that take naive C/C++ and produce executables that fully leverage RISC/EPIC platforms to be better than CISC. See also: Intel Itanium

I'm not talking about a reference check by someone other than the author. You'd not put a reference in in the first place, that you hadn't read, since you couldn't formulate the text that relates to the reference?

Ed: thanks for the link - I hadn't seen that yet.

> A human can then verify the ones with under 90% certainty.

How about the author actually reads the finished report a couple of times and checks all the references?

It really is the lowest bar - even lower maybe than running a spell check.

I guess everyone uses 20% percent of Jira - just a different 20% ... [1]

We're using GitHub for everything here, but was using Jira as an email first helpdesk.

Was hoping this was that - but apparently not at all.

We almost went with libredesk - but it's a little too simple (no merging tickets?). We're giving FreeScout a go - looks like we might need the oauth2 plugin to work with o365 mail ...

[1] https://www.joelonsoftware.com/2001/03/23/strategy-letter-iv...

> A lot of software developers are seduced by the old “80/20” rule. It seems to make a lot of sense: 80% of the people use 20% of the features. So you convince yourself that you only need to implement 20% of the features, and you can still sell 80% as many copies.

> Unfortunately, it’s never the same 20%. Everybody uses a different set of features.

-- Joel Splosky

> When I was a kid, my grandparents were involved in a pretty decent intercontinental floppy disk piracy ring. They would buy and clone software sold locally and send it forward and get copies of games in response. My parents ran a small business converting peoples university notes/recordings into well written essays. My grandparents had a PC with Prince of Persia, and as payment for my parents essay writing services one of their friends from Hong Kong used to come around and teach me how to play. See he couldn't speak or understand english very well, but he had memorised the potions you needed to drink to get past each level, and also the fighting technique of most of the bad guys.

Sounds like the summary of the opening chapter of a Bruce Sterling novel.

Love that your Hong Kong friend memorized the DRM codes.

Of course, DRM was no issue with a cracked copy of the game...

Well, there was stow/xstow...

https://www.gnu.org/software/stow/

https://xstow.sourceforge.net/

Since there was a recent thread on react compiler[r] - I wonder if adding/pushing the code through react compiler would help? (Assuming it's not already being used)

[r] Thread was about rewrite in rust, but it made me have a look at the purpose/claims made by the project - and fine-grained, automated memoization for speed seems central.

Drones might hunt down enemies running away from the target site, while a missile would only destroy the target - and those not abandoning their post?

They can both be bad.

One way in which automated drones might be considered bad, is (if) they cannot accept surrender - but are used in scenarios where human operators could.

"No quarter" is a war crime.

In addition - the docker compose example doesn't set up any data volumes for the postgres instances - that might be considered a bug?

Then again, sharding on a single host probably isn't very useful anyway - but it might work with docker in swarm mode?

Yep:

https://downdetector.ca/status/googlegemini/

Maybe take some inspiration from wap?:

https://en.wikipedia.org/wiki/Wireless_Application_Protocol

Although with supercomputers in wrist watches, I'm not sure we need a new standard - html and css works fine across devices - if used with some care?

I came across this as a Facebook memory today, and saw it had a fairly recent turn on hn a couple of years ago:

https://news.ycombinator.com/item?id=39358317

But clearly that was before LLMs captured the zeitgeist - I'd be curious to how people see it today - I'm sure a few people have parts of this (or equivalent) in their AGENT.md or similar?

I also wonder how the last part holds up:

> 80/15/5. Spend 80% of your time on low-risk/reasonable-payoff work. Spend 15% of your time on related high-risk/high-payoff work. Spend 5% of your time on things that tickle you, regardless of payoff. Teach the next generation to do your 80% job. By the time someone is ready to take over, one of your 15% experiments (or, less frequently, one of your 5% experiments) will have paid off and will become your new 80%. Repeat.

Also, reading that recent discussion on hn - it's remarkable how much LLMs and AI has changed the discourse in a short time.

I meant more like:

    0 8 * * 1-5 claude -p "Fetch my unread emails from the last 24 hours, identify urgent items, and provide a bulleted summary." >> ~/desktop/daily_email_summary.txt

(Courtesy of Gemini, not Claude code, as I'm on my phone).

Now, obviously - you might want something a little more elegant; but my point was that if you already grant Claude tool access to your email, slack etc - then it should be trivial to wrap it in a script, and run that from Cron.

I wouldn't do that; I don't trust Claude code with access to my mail (nor would I trust Claude desktop - but I don't use it anyway).

But, if you do trust Claude to read your slack and email, I don't see why Claude code couldn't do this for you almost out of the box?

While not trying to recreate the infamous dropbox comment - if you already have Claude code, are on Linux - can't Claude write a cronjob/systemd task that invokes itself for you?

So use RDP?

https://www.xrdp.org/

(Or x forwarding over ssh, or waypipie if you only need to access a gui application, not a desktop).

Interesting write up!

Have you been able to extract libraries or tools from this project yet? If so how was that experience?

That is, do you see yourself releasing a metric harness, or sub-projects that are equivalent of ActiveRecord, zod, or similar open source tooling that frequently originate in a large in-house project - and then is exported out as a stand-alone toll, utility, library or framework?

Because while ai can reimplement minor tools, it's utility entirely depends on the existence of solid tools, libraries and frameworks.

I'm using beeper - and it really lowers the mental overhead of tracking messages for me - I have some people on signal, some on messenger, some on sms, some use dms on discord.

It's great to have a single app that shows notifications and tracks conversations.

Much better than having to try and remember if my conversation with "Tom" was via sms or signal or something else; he might have a messenger account but not use the app - doesn't matter - I just open beeper and find the last conversation and continue from there.

Ed: I don't use beeper or messaging on desktop - so Franz is DOA for me.

Which is a shame - it looks like a solid product.

It's back up again - so far no comment on the blog:

https://blog.neocities.org/

You might want to look at:

https://unsloth.ai/docs/get-started/fine-tuning-llms-guide