tedmiston

Betteridge's law of headlines is an adage that states: "Any headline that ends in a question mark can be answered by the word no." It is based on the assumption that if the publishers were confident that the answer was yes, they would have presented it as an assertion; by presenting it as a question, they are not accountable for whether it is correct or not.

https://en.wikipedia.org/wiki/Betteridge%27s_law_of_headline...

you are constantly not finishing today's tasks. it lives to see another day. if you can fix it, bound it to the workflow, you will be a billionaire

dual dishwashers

Allergy immunotherapy (shots) work by gradually desensitizing the immune system for your specific allergens over several years, so yes. Some people can get to the point where they no longer need a daily antihistamine, but it's all very personalized and individual.

In any case, it's worth doing the skin prick allergy shot testing so you can have some idea of what common allergens affect you and the intensity of each.

the creative inputs to the whole process and judgment aspects of telling it how to refine and edit

Definitely. One of my fav techniques is to ask an LLM to simplify something by 90% or 99% if it looks overly complicated. Or asking the output to be critically reviewed by another agent too.

pretty big leap of faith there

likewise

The adoption curve across companies and industries is highly variable right now. Tech moves fast but plenty of boring industries need software but don't move on the bleeding edge of tech / LLMs / etc.

Even in the discourse here, you can see people getting variable quality of results and variable skepticism, some of which is valid, but a lot of it reads more like not having spent time really understanding prompt engineering.

What does a random bug in one LLM's frontend app have to do with learning how to do prompt engineering well?

What are you writing that Claude is actually writing all of it? Every time I get past the green field stage, I just end up throwing out what it writes half the time since its trash.

For the current state of frontier models, you need to break the steps down so that the LLM understands a process like what you might go through as you expect it (which is often different for everyone).

i.e., get it to agree to a spec, then get it to agree to a build plan, agree on unit test signatures, UI etc as needed, then let it build, ...

"Prompt engineering"

LLMs (at least today’s) don’t build simple solutions ...

... by default.

One area where I feel safe saying they are “wrong”, rather than just going with a different assumption that was left unsaid, would be when it makes up API endpoints. It sees the general pattern in an API, then makes up an endpoint that sounds good, follows the pattern, but isn’t actually implemented.

I remember seeing this maybe 6+ months ago, but using paid plans, RAG, and a high thinking mode has eliminated a ton (almost all) of those kinds of hallucinations. Open models and free tiers are not there yet though.

I’ve also seen a lot of issues with co-workers using an LLM to write their readme files. I look at the readme for what return values I should get, go to use them, and get an error. I check the code, and sure enough, none of the variables in the readme exist. The LLM just through they sounded good. Things like this I would say are pretty objectively wrong.

LLMs don't co-sign the quality of PRs though — your coworkers do. It's not unusual for docs to get oudated and not be maintained enough in small codebases, but that's not an LLM specific problem.

They work in that they induce taking more breaks (20/20/20 rule) but they're also annoying in that they interrupt deep focus. If the latter can become a bit smoother about not interrupting deep focus context, then they'll be more helpful.

Time Out is okay but extremely annoying in that it always seems to pop up and take over the screen at the worst possible times even with configuration to make it less interruptive. I stopped using it in favor of just closing my laptop for a bit with that.

It could be a lot better if it understood the focus context better, but it doesn't do that today. Hopefully LookAway is better at that.

I work with one guy who's just completely lost the ability consensus-build ideas with human beings. When he reaches consensus with his AI assistants on a design, analysis, plan of action, etc., he's extremely resistant to considering any other options or accepting any other input. When people do push back, he ...

To me this reads like a personal resistance to feedback / resistance to behavior change or a very jr eng more than agent related. Code isn't sacred — LLMs often easily generate absurdly overcomplicated garbage spaghetti code, but this appears to be getting better, and they are "steerable" towards better outputs with a few rounds of revisions, and some light process around a light spec, TDD, etc.

Well that's a Silicon Valley level pivot

you can do the period trick:

. http://example.com .

https://www.idownloadblog.com/2024/03/15/how-to-send-links-w...

I generalized the statement from one cloud to big public clouds categorically to show by hyperbole that ... what difference does it make?! [One can find analogous critiques of Azure, GCP, etc.]

Last year, AWS did ≥ $100B in revenue across millions of customers. But where do you draw the line exactly? "Everyone who uses < thing > is < problematic >" feels extreme.

"Anyone who uses < public cloud computing > is hypocritical" is a pretty insane take, even for HN.

All technologies have benefits and costs — choosing and using a technology does not imply the nonexistence of tradeoffs. One can give sufficient consideration to the downsides, and then determine that the upsides outweigh them. It's not rocket science.

Make Neo-Luddism Great Again! /s

I know the technology, I understand what it's doing and I know the impact, so I am vehemently anti-AI.

Author: Goes on to demonstrate superficial (mis)understanding of the technology by proliferating misconceptions peppered with anecdata and heavy virtue signaling and calling it a blog post.

Hmm, okay, then...

Is anyone else annoyed by this kind of ironically [^1] weak thinking / writing that conflates: (1) one's own personal opinions and biases however long-standing or irrelevant; (2) limited working knowledge of the actual technology; and (3) virtue signaling / moral posturing / etc? ... and then ultimately just stirs that all up in a pot to not actually say anything more substantial than "AI bad". It's such an overstated, bland, lifeless, useless, uninteresting, intellectually lazy take.

Clutching onto a weak opinion, strongly held [^2] does not make one an "outcast" ... it just comes off as closed-minded and melodramatic. Is that even contrarian? Being on the majority side of an unnuanced opinion is about as far away from being an outcast as possible...

--

Very few of the moral panic type issues those vehemently opposed to LLMs are raising repeatedly are really unique to that field... because why? [Because LLMs are not the problem.]

- Where was said moral posturing when we were building the cloud computing infrastructure?

- Where is the concern of "wasting" compute resources when using 10–15 GB of bandwidth to stream a 90-min movie in 4k?

--

[^1]: Better not call the poorly written human authored post of poor quality "slop" though!

[^2]: Not a typo.

Reads like a post straight from r/iamverysmart.

I’m terrified that there’s going to be a doctor’s office sometime that does the same, with more serious consequences.

they can send scheduling info, appt reminders, etc via SMS but (1) they must allow opt out, and (2) they cannot send medical info this way — that's where HIPAA requires encrypted "patient portal" messaging because SMS can be intercepted or accessed by others.

From my quick research online, it seems they've gone digital-only for season tickets because they don't want people just reselling them to turn a profit. They want actual season-long fans, so now if you transfer too many games they can track it and ban you.

this is a common clause in season-ticket memberships, but it doesn't actually work all that well. for instance, resale on the ticket marketplace is tracked and counts, but in general transfers alone are not penalized. so people do transactions outside of the official platforms, sell / trade in fb groups, etc.

RatGPT

And as for the critics, tell me I don't get it

Everybody can tell you how to do it, they never did it

—jay-z

sure, but undo isn't the only path to a newer better version of the code

it's easy to see how the product (claude code) could be abstracted to spec form and then a future version built from that without inheriting previous iterations tech debt

ehh, as long as the overall starting architecture is decent, it's not hard to do tiny refactors across components

claude code, the app, is also not some radically complex concept (even if the codebase today is complicated)

but hey, that's why people do version breaking rewrites

they prioritize bugs based on what impacts the largest amount ($) of customers — fixing every bug of a huge complex project rarely makes sense if the impact of the bug on revenue is tiny

i think you are conflating anthropic (the startup) with claude code (the leaked source of one of said startup's products)

i.e., the claude code codebase doesn't need to be good right now [^1] — so i don't think the assumption that this is an exemplary product / artifact of expert agentic coding actually holds up here specifically

[^1]: the startup graveyard is full of dead startups with good code