dmarble

geçen yıldiscuss

Direct link to the spec page on authorization: https://modelcontextprotocol.io/specification/2025-03-26/bas...

Source: https://github.com/modelcontextprotocol/modelcontextprotocol...

dmarble

2 yıl öncediscuss

Plot twist: The * in C-00000291*.sys is "-block-ultron"

Premature deployment of Crowdstrike AGI disaster response plan.

dmarble

2 yıl öncediscuss

I'm in awe at the complexity and unknowability of it all, but I also have to chuckle at the thought that some portion may be vestigial.

I'm particularly fond of the "Egg shaped object with no associated processes". :)

dmarble

3 yıl öncediscuss

Very similar story here on the PM side in the US.

~15 YOE (FB/Meta most recently), ramping up a search after a sabbatical, targeted job search to roles where I have non-trivial experience and domain expertise, customized cover letters, leveraging my network, open to relocation, open to hybrid or remote, etc.

I'm seeing 3-5% response rate over the past few months. It's rough out there. No response for seemingly great matches. Slow moving recruiting process, even at early-stage startups. Rejections after screenings and first round interviews where the mutual fit seemed excellent.

Hiring manager friends and talkative recruiters tell me that in contrast to the past decade where they'd routinely screen people who met most of what they're looking for, they're now dealing with a massive volume of very qualified applicants (and trudging through a massive volume of unqualified applicants). Deciding who to screen and who to do a first round of interviews with is taking a lot more time and effort. And the first round of interviews might include 6-8 unicorn (i.e. perfect) candidates, where in the past they'd be elated to find 1 unicorn.

I've been through a couple down cycles, so I'm focused on grinding away till I find something. I think every level is feeling the pain in some proportional way. Big sympathy for early career folks. Even if we reset compensation expectations, it'd be a shame if the sector ends up losing out permanently on a range of talent. There's no way tech needs are going to decrease on a medium-term horizon (though they may shift).

dmarble

3 yıl öncediscuss

Check out:

NocoDB (https://github.com/nocodb/nocodb) - works w/ MySQL, PostgreSQL, SQL Server, SQLite, MariaDB

Baserow (https://gitlab.com/bramw/baserow) - PostgreSQL based

APITable (https://github.com/apitable/apitable) - MySQL based

dmarble

4 yıl öncediscuss

India's beef exports are primarily buffalo meat ("carabeef"), not cow meat. In addition, there are states in India where slaughtering cows is permitted.

dmarble

4 yıl öncediscuss

Obsidian (https://obsidian.md/) + a few plugins (enable on the "Community plugins" page in settings):

- Advanced Tables (handful of features including sort col while in edit mode): https://github.com/tgrosinger/advanced-tables-obsidian

- Sortable (Wikipedia-like col sort for view mode): https://github.com/alexandru-dinu/obsidian-sortable/issues

- Outliner: https://github.com/vslinko/obsidian-outliner

- Auto Link Title: https://github.com/zolrath/obsidian-auto-link-title

For even more fun:

- DataView: https://github.com/blacksmithgu/obsidian-dataview

dmarble

5 yıl öncediscuss

Check out Legend (formerly moo.do): https://legendapp.com/

Found it several years ago and used it regularly for a while. It has a nice mixed model for blending calendar, email, lists, tasks. Lots of drag and drop capability, keyboard shortcuts. Loads fast. Mobile app. Customizable views you can quickly switch between, e.g. 3 column email on left, tasks in the middle, calendar on the right.

dmarble

5 yıl öncediscuss

Huge congrats to you and the crew, Spencer! Well deserved. It was always a pleasure working with your product and the people behind it at prior companies, and lobbying for more adoption as a result.

1. What are you most looking forward to (product/tech-wise, impact, financially for you, financially for the organization, etc.) that may not have been possible just a few years ago?

2. What do think are the greatest challenges you will have to face in your role as a founder-CEO and as a company in AMPL's next phase?

3. If you were starting over today, what ideas might get you excited enough to go at it again for the next decade?

dmarble

6 yıl öncediscuss

Don't any of the major commercial open source vendors offer custom terms to give access to the commercial source? I'd imagine they'd contemplate it for big deals. Seems like one of the only ways to keep some of these sophisticated customers onboard.

dmarble

6 yıl öncediscuss

I work on identity resolution problems among other data challenges at a big retailer. We get very few details when a credit card is used in-store -- pretty much just the basic card info. We don't own the payment networks and a variety of legal and business reasons prevent us from doing what may seem to be possible to get an individual's data based on a card number (from the networks themselves or other data brokers). What's possible is mostly limited to what you can do with the name on the card and store location. We've invested significantly in the problem and the answers are a patchwork of guesses.

There's an interesting related issue here for brick & mortar businesses with CCPA and GDPR in effect: you can do some useful analytics, personalization, and fraud prevention work with probabilistic identity info, but if someone verifies they actually are Person X and wants to download or delete whatever data you have on them, what can you confidently say is actually their data?

Will companies be held to different standards based on how much money they've invested and success they've had in identity resolution, in which case this might be a factor dissuading them from doing more identification and personalization? Or if they haven't invested millions in trying to figure out who people are, but it's possible to do so, are they liable for some kind of misconduct if they don't produce all the data they have that could have been tied together for that person? Is the choice binary, i.e. either invest big in identity resolution and take it as far as possible (with parallel governance investment) or de-identify everything you can? A privacy advocate might think on first pass that it's as simple as choosing the latter, but that's mostly not possible due to requirements we face related to other regulation and business realities: fraud, anti-money laundering, age-related laws, shoplifting, intense competition in a razor thin margin industry, etc.

Data privacy is complicated.

dmarble

7 yıl öncediscuss

  On Android Chrome with Data Saver turned on, elements with loading="auto" or 
  unset will also be lazily loaded if Chrome determines them to be good 
  candidates for lazy loading (according to heuristics).

I don't agree with this design decision to make the default ("or unset...") allow for lazy loading, even if limited to Data Saver enabled phones. Doesn't this mean every site on the planet that deems Android >= 7.0 web traffic important and makes use of pixel-based or iframe-enclosed tracking will have to go through testing and potential modification?

dmarble

11 yıl öncediscuss

"There's a good half-century of research around high expertise development but some of the most powerful and interesting has been greatly underestimated. The short version: modeling, modeling, modeling. If you want to develop high expertise, spend time, lots of time, observing those with high expertise. If you want to help others develop high, deep expertise, put them in a context where they have as much exposure as possible to expert examples. This could be either examples of the work experts have done or, depending on the domain, observing the experts as they do their work. The problem is, we let beginners (and intermediates) spend way too much time being around mostly other people who are also beginners. In the ideal, perfect, not practical world... everyone would be surrounded only by those who are not just a little better, but with very high expertise.

"The only catch — but this is a tough one for most of us — is that this works far better if the expert does not try to explain what they're doing as they're doing it. In most contexts, the learning is much more robust and accurate and deep if the learner just... watches. Without trying to consciously analyze or ask for justification for every little decision. There’s a time for analyzing the work of those with more expertise, but the counter-intuitive research shows quite strongly that this is more likely to reduce performance, of both the learner and—in many cases—the expert trying to explain how they do what they do!"

Source: http://www.cioinsight.com/it-management/expert-voices/kathy-...

It'd be supremely awesome if companies like LiveCoding could convince experts to live code from time to time. Would certainly pull in lots of visitors. Might even be able to charge for that. Or get experts to record themselves in the flow of work without a live audience, and make a library of such recordings available. Might be a great addition to some EdTech company's offerings.

dmarble

11 yıl öncediscuss

A couple summary quotes:

"[W]hen [Dr. Kent] Berridge, a dedicated young scientist who was more David than Goliath, stumbled upon evidence in 1986 that dopamine did not produce pleasure, but in fact desire, he kept quiet. It wasn't until the early 1990s, after rigorous research, that he felt bold enough to go public with his new thesis. The reward system, he then asserted, has two distinct elements: wanting and liking (or desire and pleasure). While dopamine makes us want, the liking part comes from opioids and also endocannabinoids (a version of marijuana produced in the brain), which paint a 'gloss of pleasure', as Berridge puts it, on good experiences. For years, his thesis was contested, and only now is it gaining mainstream acceptance.

. . . .

"'It's easy to turn on intense wanting,' he says, when we eventually sit down. 'Massive, robust systems do it. They can come on with the pleasure, they can come on without the pleasure, they don't care. It's tricky to turn on the pleasure.' He hadn't expected his findings to turn out this way, but it made sense. 'This may explain', he later tells his audience, 'why life's intense pleasures are less frequent and less sustained than intense desires.'"

== On mindfulness meditation: ==

"It's not that meditation makes the wanting go away. 'What it is doing', Berridge says, 'is giving the more cognitive mind a way of distancing itself from the urgency of those wants. It's a practised mental gymnastic. A want occurs, but because you're so practised, you can recognise that want, appraise it, feel it all around, focus on that, and the feeling of urgency as a feeling, without engaging in it.'"

== On controlling our desire/pleasure systems: ==

"The strength of dopamine/wanting in an addict's brain is so fierce that it is hard to conquer. ... Walter Mischel's famous marshmallow tests told children that they could forgo one marshmallow for the promise of two if they waited a while. Mischel tracked the children in later life and found a link between self-control and success. The controlled kids had resisted the marshmallow by simply making a decision and moving on without further discussion. They turned away from it, or tugged their pigtails to distract themselves from allowing it to arouse their senses. The children who deliberated, or lingered over the marshmallow, were more likely to cave in. 'It looks as though the best way of resisting is not to open the question[.]'"

dmarble

12 yıl öncediscuss

Xtrafinder has been my saving grace, and it's free (https://www.trankynam.com/xtrafinder/)

- Cut & paste

- Hide dotfiles on Desktop

- Automatically adjust column widths

- Backspace to go to back

- Add "go up" to toolbar

- Add functionality to menus (e.g. new terminal here, new file (with templates), copy path, show hidden items, refresh, launch as root)

- Dual window

- Visual tweaks

- A few other niceties

dmarble

12 yıl öncediscuss

Good breakdown on the Sencha blog, with some benchmarks and a discussion of the few major bugs that prevent replacing UIWebView at this time: http://blog.sencha.com/blog/apple-shows-love-for-html5-with-...

dmarble

12 yıl öncediscuss

Georgetown law grad here who maxed out loans, worked during school, and put the excess capital into my first startup (cheaper than a small business loan!), which was law/tech-related. Not that I recommend that approach... I'm now back in software product management, which is what I did before law school.

I tell people law school is a bad book club. You read a crap-ton of mostly boring books, talk about them, and usually take a test at the end of each course. Occasionally there's an amazing instructor to make a subject enjoyable.

I'd prefer in the US we keep law as a graduate program, but tear down the antiquated 3 year system. Law school should be one year, with optional "residency" and specialist training like medicine. You'd get the research, writing, and basic subjects in that one year, which are all you need for most actual lawyering. Residency would be the hands-on training you never get at law school unless you had an amazing clinical program. Bar exams should be 1 basic exam covering that 1-year curriculum, and additional exams if you want to be certified to practice a specialty.

dmarble

12 yıl öncediscuss

Very cool. Good luck to you and look forward to checking it out!

dmarble

12 yıl öncediscuss

If you haven't taken a look at Nimrod, it might give you some ideas, or perhaps inspire you to be a code contributor! As a python guy, I'm really liking it for some hobby work I'm doing. The compiler and standard library are MIT licensed.

http://nimrod-lang.org/ https://github.com/Araq/Nimrod

Also, there's a recent forum posting about the decision not to do LLVM codgen for now: http://forum.nimrod-lang.org/t/480. Perhaps you can ask some questions of the main contributors; they're quite responsive.

dmarble

12 yıl öncediscuss

Wow, quite the list! Now I feel compelled to read through some posts tagged "AirBnB for X".

Interesting discussion on whether this model can succeed: http://www.quora.com/Will-Airbnb-for-food-be-a-successful-bu...

And I just stumbled onto another one of these last week: http://cozymeal.com/

Submissions

Rails has Omniauth. Django has... a ton of one-off, limited social auth apps?

comments