Hacker Newsnew | past | comments | ask | show | jobs | submit | jll29's commentslogin

Lab diaries are also very helpful to track the provenance of ideas/inventions.

This reminds me: when people insist on having a real phone call in an email, it could be something that they don't like to put in writing. So it's a good practice to ask what the topic of the phone call will be so that you can join it prepared.

If it indeed is something that you feel might be fishy, I further recommend the following: write a summary of what was discussed and send your summary to the people on the call as "meeting minutes -- 2026-02-11" (make this a habit, and always say "I do this routinely to remember what was agreed"). This can easily avoid you being trapped by dubious propsals or being unwittingly on the wrong side of the law.


You can ask at the beginning of the call if it can be recorded.

I record my phone calls for personal records. Often I won't hear or remember details, so the recording helps.

Are there legal or other situations in which meeting minutes would be admissible in evidence, but a recording would not be? Obviously this is jurisdiction dependent.


Which phone and which software?

dogs see violet better, so "normally" our sky would be blue to them. But because their eyes have only two types of color receptor, they see violet as blue, and our sky is also blue to them.

Same outcome, but for a different reason!


Although dogs can see violet better than humans, the sky appears blue to them like it does to us, not violet (despite violet being closer to the resonance frequency of N and O atoms).

That's because dogs only have two types of cones (color receptors) in their eyes, blue and yellow, not three as we do, so they see violet as blue.

Therefore, the sky is blue to us and to dogs, but for different reasons!


Great app, pleased to be in the top 0.38%, but it appears that does not translate to a top 100 spot by an order of magnitude.

It would be nice to have some readership stats, too.

I've been wondering whether Webcam-based eyetracking software could be used to calculate via triangulation/trilateration which word one is reading on the screen.

Then words could be color-coded by impact.


> Then words could be color-coded by impact.

Or by difficulty to make sense of them.


I am in the top 0.15%. Crazy.

Top 0.04% here. Apparently I could have written 1.97 volumes of game of thrones with my HN comments (590 000 words). I don't know whether to be proud or embarrassed. I think I'm a little of both.

Let's see:

  do {
    4 h HN;
    4 h family;
    4 h coding;
    4 h sleep;
    4 h emails;
    4 h meetings;
    4 h teaching/supervisions;
    4 h grant/paper writing;
    4 h reading;
  }
  while (breathing);
  //
  // ----------
  // 36 h / day
  // ==========
But only some of the meetings and grant writing feels like real "work"; the rest feels more like "I'm getting paid for my hobby" (including for reading HN, no kidding), and if it wasn't my job I would be paying money to be allowed to do it.

How is live with 36h days?

Installing the binary on a Mac succeeds, but then:

% msgvault init-db dyld: Symbol not found: _SecTrustCopyCertificateChain Referenced from: /usr/local/bin/msgvault (which was built for Mac OS X 15.0) Expected in: /System/Library/Frameworks/Security.framework/Versions/A/Security

zsh: abort msgvault init-db % msgvault add-account ****@gmail.com dyld: Symbol not found: _SecTrustCopyCertificateChain Referenced from: /usr/local/bin/msgvault (which was built for Mac OS X 15.0) Expected in: /System/Library/Frameworks/Security.framework/Versions/A/Security

zsh: abort msgvault add-account ****@gmail.com


Currently, all hyperscalers are in an arms race to buy energy, data centers and GPU capacity for the big struggle of "who will win the AI war?" - the Amazons, Googles, Oracles, Metas and Microsofts of this world all want to be the one winner that takes it all.

In this case "it" or rather "they" are the consumer AI market (a la ChatGPT) and the B2B/corporate AI market (for enterprises and their vertical AI applications) - and yet to be built!

The race is run before we get to know who the killer apps are, and whether there will be one (if not: another bubble to burst, much worse this time than in 2001).

Perhaps the XAI/SpaceX integration could be done for the same reason that Amazon fired 30k white collar workers: enormous amounts of cash is needed for the investments in AI infrastructure.

Now XAI alone may struggle to raise that much; with SpaceX integrated, there are government funds to tap into as well, and it's less clear to separate the space travel business from the AI business, which might be intentional. Amazon shedding 30k people means approx. $6 bn in annual savings, capex that can be redeployed (cf. https://www.youtube.com/watch?v=7sk3qmIQZnI)!

EDIT: Typo fixed


My empirical findings (30 year time frame, dozens of projects, a few larger, some small many small-to-medium, median perhaps 18 calendar months with around four people on the core team):

1. Good estimates are possible.

- use intervals, not single numbers; intevals are asymmetric (worst case is far out compared to best and expected cases);

- calculate using PERT estimates;

- estimate by combining top-down and bottom-up estimates; when both agree, trust the estimate; otherwise, refine/revise the plan;

- let the assignee estimate their own work, but the manager correct it (because of (2));

2. Better engineers tend to underestimate their own time more than average performers.

If management tells you how long you got, look surprised and ask them how they can know that; then tell them you will now working on calculating how long it will really take. If their "gut instinct" and your professional estimate diverge, show them your plan, talk them through it, and ask them which part you should leave out (if you keep the time fixed, perhaps you can shrink the scope).


It's a very clever aspiration to devise a new language not as something you hope everyone is going to switch to, but, as the OP states more of a test-bed to demonstrate a bunch of nice features, which you hope other people (that implement mainstream languages) will borrow/steal/copy. For instance, I very much appreciate the automatic parsing of command line arguments (and beyond just strings), which I hope the Rust folks will take over one day. Who would not like to skip all the boiler plate writing, but still offer decent cmd line options? For that reason, I will not compare the current Tomo feature set with any other language (as many other commenters do). But I will say that 150 lines for a complete terminal "snakes" game is pretty cool!

It's also smart to facilitate integration with C or other languages that have an abundance of libraries, because it's unlikely that you will create the momentum to rewrite everything in your facorite baby language.


For CLI arguments, have you checked out clap? It's declarative (you create and annotate a struct, it generates the parser), and can be agremented with man page generation or shell completion generation.

And as a result of the parsing step, you get a fully typed struct


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: