HN user

mabcat

492 karma

e:hn-at-e.mab.pm lemmy:@mabcat@programming.dev https://programming.dev/u/mabcat

former project: OkToTalk, recommender and matching system to find you the right psychologist for your situation and personality

If you've ever wondered why everyone says "therapy is great, you should get a therapist" and yet it's ridiculously hard to actually do this, I can discuss it at length...

Posts0
Comments86
View on HN
No posts found.

It's often said that to be outstanding, you should master two different fields then work at the intersection of those fields. For a lot of people the model of "real greatness" is Leonardo, who mastered about 7 fields.

To even get close to that you need to be the sort of person who changes fields every 5 years or so. I think this temperament is relatively fixed, perhaps because it's a feature of some kinds of neurodivergence. If ND and innate ability are both fixed, what you need to reach greatness is a bag of tricks like the author has so you optimise the many parts of personality and behaviour that are not fixed.

What I think is sad about modern times is, it's increasingly difficult and impractical to make a career change into a regulated profession. You can switch to CS, painting, or poker any old time. But if you're not a doctor by the time you're 40, forget about it. I bump into many people who are interested in medicine and would be great at it, but the switching cost is insurmountable. A shame because there's so much greatness to unlock at the intersection of CS, medical research, and clinical practice.

Closest name I can think of is "commitment and consistency". People tend to behave as they have behaved in the past, doing so is both a cognitive shortcut and a source of positive emotion. We go to great lengths to maintain consistency (see also: confirmation bias), and being consistent even in the face of conflicting evidence feels better than being inconsistent but right.

From Cialdini: "Once we have made a choice or taken a stand, we will encounter personal and interpersonal pressures to behave consistently with that commitment. Those pressures will cause us to respond in ways that justify our earlier decision."

VB was probably the charmingest language I’ve ever used, especially for GUI. It takes the cake for “rather complex runtime” though. As I recall, they bundled the entire VB interpreter into every executable. This seemed scandalous to me at the time.

Agree. Shimmer is currently in the "assume good faith" category and it would help everyone if commenters gave feedback and asked questions from this perspective.

From what I've seen of Shimmer, the rapid-fire negativity in this discussion is not warranted. I have run a mental health startup. Criticism is constant and highly energised, and it really wears you down after a while. I expect they're also copping it from established practitioners, their industry bodies, and various regulators.

I currently use Shimmer, I have used BetterHelp and TalkSpace. These last two are... not in the "assume good faith" category. Perhaps they could show up to HN for some "robust discussion".

The thing I would like from Shimmer is a policy and mechanism where if you stop accessing the app and showing up to coaching, they stop billing you. Of course this would be startup poison, businesses are built on subscription revenue from non-users. But only billing for care you provide is the strongest show of ethics and consumer-centering I can imagine. Especially when your client base is ADHDers! Shimmer, if you currently do this, put it on the front page in bold.

I know a nuclear physicist who got paid to go through the scripts and turn "CHARACTER: [insert science thing here]" into dialogue

This looks excellent! It's a great interface for two things I'm struggling to make LlamaIndex do: explain and debug multi-step responses for agent flows, and cache queries aggressively. If I can work out how to hook it into my LlamaIndex-based pile, happy days.

Feature/guidance request: how to actually call functions, how to loop on responses to resolve multiple function calls. I've managed to mock a response to get_current_weather using this contraption: https://pasteboard.co/aO9BmHG5qsFt.png . But it's messy and I can't see a way to actually evaluate function calls. And if I involve the Chat Turn node, the message sequences seem to get tangled with each other. Probably I'm holding it wrong!

I think that's the wrong end of the stick. When you publish research, the software you used/built on is part of the methods and needs to be cited. The authors are doing you a courtesy by providing a pasteable citation.

Similar "we would appreciate citations" statement for (BSD-licensed) pandas: https://pandas.pydata.org/about/citing.html

8000+ pubs citing pandas: https://scholar.google.com/scholar?cites=9876954816936339312

It's an alternative, does a similar job, depends on/abstracts over langchain for some things. It's easier to use than langchain and you'll probably get moving much faster.

They've aimed to make a framework that starts concise and simple, has useful defaults, then lets you adjust or replace specific parts of the overall "answer questions based on a vectorized document collection" workflow as needed.

This works well overall, but some bits have kept me scratching my head for hours. Partly due to huge holes in the documentation when it comes to specifics ("plenty of examples but little documentation" another commenter wrote and I agree). Partly due to the frenetic release schedule, this project is highly active even by frothy LLM craze standards and interfaces change rapidly.

Overall recommend, LlamaIndex has helped me make good progress on my project

My situation somewhat overlaps with yours. I ordered a "water flosser" from everyone's favourite ecommerce site, and use it daily in the shower.

It seems to be doing something, and can't possibly be worse than my previous flossing regime which was: "twice a year the day before a dentist visit".

I wasn't willing to shell out for the name-brand water flosser, other options do the same job at the same water pressure for 1/4 the cost.

It took about a month to move up through the power settings. Start on the lowest and click it up when that starts to feel a bit tame.

I've now been using this thing for 6 months, it's just about time to find out whether I get my bi-annual nagging about the importance of flossing. My personal opinion: Cochrane's probably right, flossing is most likely bunk, but I'm pleased to have found a way to do it that I can stand

You do need to separate medium of exchange from store of value. Real estate is useless as a currency while being the primary store of value for most people. Most of your objections also apply to real estate: can't know how much wheat a house is worth, houses are notoriously illiquid, price discovery is difficult and quite opaque, the sector is rife with misrepresentations. I wouldn't go so far as to say fraud, but where does "real estate agent" sit on those polls of "most trusted professions"? Houses aren't even fungible!

And if anyone thinks BTC/ETH transactions are slow and expensive (me!), around here the gas fee on real estate transactions is ~5% with a 6-month processing time...

From the study, in-person schooling increases teen suicides by 12-18%. Is this fatality rate acceptable on the basis that it's "very necessary" to socialize the remaining children in this way? Is there evidence that kids in online classes wouldn't have "grown up"?

I find this line of thinking abhorrent. If it will reduce suicides by 15%, in-person schooling should be stopped outright until someone fixes what's wrong with it.

masto.host was an option yesterday, now it looks like it won't be an option for the next few days.

I was about to install self-hosted Mastodon but the machine load and amount of admin chores seemed too high. I've installed self-hosted Pleroma instead. Pleroma has 2% the DAU count of Mastodon, I'm starting to wonder if I'm missing out on the main experience, but I also don't want the main experience to be "server administration".

That's such a strange piece of UX. As a typical user, who on average consumes content rather than produces it, what I care about is that my list of followed accounts migrates with me so that my feed stays relatively consistent. As a content producer I would also care about my past posts migrating with me, which they don't. Having followers continue to follow me after migration is only the primary concern for the "influencer". I get that "my friends can still find me" is an important feature but who thought it was ok to skip the other two?

It looks like the Mastodon instance host Simon is using, masto.host, closed for new sign-ups 8 hours ago. Is there another recommended alternative? I don't think I have the energy to maintain a self-hosted Mastodon instance. I sure haven't missed running a mail server...

Stability is how the aircraft reacts to disturbances (control inputs, turbulence, ...). Stable aircraft resist disturbances and settle into going where they're pointed at a predictable speed, so you can let go of the controls and attend to other pilot things. Unstable aircraft, when disturbed, go more and more in the direction of the disturbance. (e.g. a turn gets tighter and tighter, nose oscillates up and down more and more, ...) You need to correct this with more control inputs, but these will also have unstable results. So they only let computers fly unstable aircraft.

A car has good stability in turns: it wants to go straight, if there's a bump or you jerk the wheel a bit it still goes straight. If you apply a control input and get a turn, it will let you maintain the turn but it also wants to unwind and go straight. This is stability but not aerodynamic stability.

The reason one might want a less-stable aircraft is that stability is the opposite of manoeuverability. Super-stable aircraft like training gliders really resist turning and can feel too much like hard work.

As another commenter mentioned the wings don't contribute much to stability. The two big fins at the back (the stabilizers) are the main source.

(disclaimer: I fly fixed-wing acft and gliders, barely, and am no physicist.)

I'm familiar with the industry. Everyone involved knows that you don't move so much as a mobile phone's worth of lithium without following a procedure. In my experience small violations are wilful laziness, large violations are to avoid the signficant costs involved with dangerous goods documentation, packaging, and handling. This container didn't get onto the truck without several people deciding not to do things they're trained and certified to do.

[dead] 5 years ago

I tried one of those, it didn't make a shred of difference. MBP 16" still throttled so hard it was frequently unusable.

If you're in Australia, find a therapist using the thing I built: https://www.oktotalk.com.au . If you fill out the questions, the site might be able to recommend you someone good straight off. Otherwise one of our staff will help you through the process. We try really hard to bypass that "I tried 4 therapists before I found the right one" thing, and get you in with someone right the first time.

Here's the 80/20 of getting the most out of a therapist, drawn from watching thousands of people get great results over the last 5 years:

* talk to several therapists on the phone for 5 or 10 minutes first. You're going to be hesitant and uncomfortable, it's their job to put you at ease, they're the professional. You'll learn a lot quickly about them, and about the mental health process. If you have an uncomfortable 5 minute chat, you're going to have an uncomfortable hour, so skip that therapist. Go with your intuition, either say "heck yes" or "no".

* make a plan, work the plan. Write out your goals. The way to do this is "present situation" -> "desired situation". Everyone forgets the "desired situation" because they've got so much negative mental energy tied up in "present situation". Anyway, take your goals to your therapist. Put a circle around the one you need to address first. Insist that it gets addressed, or 6 sessions will go by and you'll realise the only change is to your wallet

* ask for homework, do your homework. You're in therapy an hour a week and out of therapy 167 hours a week. Your therapist can't compete with that. What you do outside therapy matters greatly

* if after 3 sessions you haven't established a good working relationship with your therapist and seen a bit of forward motion on your goals, get a new therapist. It could improve but statistically it's unlikely. This is an evidence-based cut-off.

Mental health is health, stay healthy!

I do the separate-email thing too and my spambox is chock full of emails to leaked addresses. The leaks seem mostly from large, upstanding companies that have had data breaches (Adobe, Myspace, ...), rather than from deliberate email address sales. It's still very handy to be able to just disable an email address... can't do that if you've used the same registration email on 1000 sites.

I find that Airpods Non-Pro will stay in when I'm on my side and when I roll over. I'm a terrible sleeper and change position a fair bit, but there's no problem falling asleep with them in. I often wake up with them both still in.

any resource recommendations or tips for writing a translation engine

All the successful translation websites/apps you're familiar with use machine learning. ML stomped all over NLP approaches because it gives a rough translation between so many more languages for so much less work.

On the question of where the data comes from, you might be a bit closer than you think. That dictionary you're transcribing has some sentence pairs, and like yorwba said, sentence pairs are food for ML language models. Extracting all the sentence pairs into a dataset might raise some interest from ML people.

In our case we've already got all the stuff for serving web pages out of Django... the data model, DRF, auto scaling, access control, CI/CD. If I can use the same infrastructure to serve predictions, without having to duplicate anything or learn new frameworks, that's a huge win. Dealing with CPU overhead is easier than dealing with brain overhead.

End-to-end encryption has been named as a required feature for telehealth in Australia. Interest in telehealth has gone from zero to infinity over the past two weeks for obvious reasons. So I've been trying really hard to work out if Zoom is E2E, and reached the same conclusions as the article. First, it isn't, and second, Zoom are really going out of their way to obscure that fact.

It's great that The Intercept is taking a look at this, because it's absolutely beyond the capabilities of healthcare practitioners and the professional bodies to get to the bottom of. There's a ridiculous amount of confusion here, compounded by "you need to get the HIPAA version because HIPAA means privacy".

We ran a clinical trial of a very common antidepressant. Blinding was difficult because on the active arm damn near everyone had observable side-effects within 40 minutes of administration. Clearly not related to the placebo effect in this case, only the pharmacist knew what was in the capsules until the nausea started.

The charge controller doesn't allow simultaneous charge and discharge. It might allow simultaneous charge and power. The concern is, if your power pack does let you power devices while charging the battery, it's passing the power straight through from the wall, not via the battery. Is it also passing the data lines straight through? Not sure, you'd have to check the schematics. Here's an example... a great board, I own one, but I have no idea if the data lines are isolated: https://www.adafruit.com/product/2465