HNHacker News
TopNewBestAskShowJobs

halfjoking

567 karma · joined July 9, 2017

submissionscomments
halfjoking··on Tell HN: Cloudflare verification is breaking the internet
Imgur links haven't worked on my VPN for a long time.

Even if they did, I'd still avoid imgur since they censor even worse than reddit.

halfjoking··on Ask HN: Anyone else see GPT-4 access as unfair?
Yes - it should be everyone gets access, and they rate limit based on how much demand they could support. If that means everyone only gets 3 requests per day for GPT4, then so be it. That would be fair.

It's disappointing since Sama was a part of YC which does seem to care about merit and "fairness." So you'd hope he'd support the same principals to "support the little guy" and give everyone a shot at having the best tools.

At least GPT4 isn't good enough to use something like AutoGPT to actually be useful at self-developing large portions of software. Once AI is that good, API access will be critical. Entire infrastructure and first-mover advantage could be dependent on API access at that time... there would be AI pipelines that wouldn't have been possible months ago that a highly intelligent AI model could see to use instantly and build out in a day. Imagine if your competitor has access and you don't in that AI future.

halfjoking··on Verified accounts are now prioritized
I think it’s funny Elon is gifting famous people the blue check. PG suggested Elon should allow any user to pay for another’s check.

I hope Elon does it so people can give the gift of shame.

halfjoking··on Longevity study across 5 species found a new pathway to reverse aging
Get good at fasting. Caloric restriction is the only thing proven to increase longevity, other than basic things like healthy eating and exercise.
halfjoking··on Methane may not warm the Earth quite as much as previously thought
There are a shitload of climate deniers.

Maybe you never heard that Bill Gates is going to have “climate lockdowns” after covid, or “control the food supply.” But among half the voting population this is a common belief.

halfjoking··on Substack is the biggest threat to newsletters ever
This is the reason I never paid for any substack subscription.

There are about 15 authors I'm willing to support in some way, but not $5/month for any of them individually. I don't even mind if they kept some articles restricted. It could be that I can "unlock" 10 premium articles per month for $10/mon, the main thing is I want to control what I contribute. $75/mon is too much.

halfjoking··on Southwest Airlines halts flight departures amid technology issue
That means you can upsell them after they deploy. Have the bot read back the formalized updates to confirm if you want more accuracy.

And don't forget premium voices. I'd argue using General Adama's voice from Battlestar Galactica leads to less mistakes.

This is Battlestar actual, what's your status?

Who won't give accurate information when they hear that?

halfjoking··on Southwest Airlines halts flight departures amid technology issue
Grandparent should pitch he can fix it in 2 weeks for $1 million.

Use vocode.dev to take the pilot calls. Use bot to ask the relevant questions, and get transcriptions. Send transcriptions to ChatGPT to formalize into well-defined json updates to the database. It's a 2 week project, max.

halfjoking··on Monthly Fediverse posts cross 1 billion for the first time
If you’re on HN long enough you should know of a superior protocol called Nostr that doesn’t tie your identity to Federated servers.

By tying each individual to a server you are allowing cancel by association culture. Individuals should not be put in left/right buckets (or any other buckets) and have that association be used to censor/exclude them. The structure of the Fediverse encourages in/out groups. That’s why it sucks, and that’s why giving individuals the power over their own identity will be the future of social.

halfjoking··on U.S. may change how it monitors the web after missing leaked documents for weeks
Main thing they learned is to control both sides of the narrative.

Make the “wrong side” look like flat-earther types through coordinated information poisoning. That’s the true advancement in PsyOps in the last 10 years.

halfjoking··on Show HN: GPT-4-powered web searches for developers
I signed up to the GPT4 waitlist the first day and I still don't have access.

As these AI models get more powerful, giving certain people months to use them while the rest of us twiddle our thumbs seems unfair. It should be everyone has access but you only get X API requests per day. Then increase X for all users evenly. If OpenAI isn't going to be open, at least they could be a little more fair with access.

halfjoking··on Even when we know they’re “fake,” placebos can tame our emotional distress
I don’t agree that the nocebo effect can be tested in a clinical trial which only has willing participants.

If there was a person who’d rather take a bullet than the vaccine, don’t you think the nocebo affect would be stronger? My feeling is anyone who requests an exemption due to their beliefs should be given it because you can’t measure the resoluteness of their beliefs, religious or not.

halfjoking··on Even when we know they’re “fake,” placebos can tame our emotional distress
Placebos can also work in reverse - if you strongly believe a medication or procedure is harmful then you will be harmed by it.

That's why it's entirely unethical to force a medical product on a population.

Take for example the mRNA covid vaccine. Let's say it did what they claimed at the start and stopped transmission 100% without ever causing a side effect. Even then, the emotional distress and reverse placebo effect of coercing people to take a vaccine if they don't want it still outweighs any benefit.

People who believe a vaccine is dangerous but take it to save their job will get harmed whether it caused a side effect or not. The emotional distress alone can be debilitating. If the mind is so powerful it can strongly affect health outcomes, how could anyone think it's ethical to punish someone for making their own medical decision?

halfjoking··on Twitter is now blocking Matt Taibbi from search results
Taibbi has been consistent with people who support free speech. I don’t understand the “you get what you deserve” mentality, when there is nothing he could have done differently to help expose government censorship.

You call out government being fascist censors, and you call out power-trip billionaires when they do it too.

It’s not like stupidly playing with leopards and having your face eaten. It’s more like being a journalist in a war zone, reporting war crimes from both sides, then being treated like a criminal by both.

halfjoking··on Show HN: A ChatGPT TUI with custom bots
Yeah me too - all I can do now is build around gpt-3.5-turbo and assume the responses will be similar to what I get with my GPT-4 with my Plus membership.

You’d think since I pay for Plus and API credits I’d get access to GPT-4 but nope.

halfjoking··on I wish GPT4 had never happened
They aren’t trained on assembly. Even if they were, the tools for debugging assembly wouldn’t be as helpful if the LLM wanted to iterate when given compiler errors. The compiler errors they’re trained on are errors humans run into.

Now they also have training on frameworks but those keep changing. So I think the best abstraction is plain javascript - it has the most training data and javascript written twenty years ago still works today.

halfjoking··on I wish GPT4 had never happened
I’ve also been wondering if no framework coding will be more productive for LLM driven software.

How many times have frameworks like React changed since it was created? A changing framework means the LLM may mix training from multiple versions and create errors. Whereas libraries that don’t change can produce consistent results. So use stable libraries that choose backwards compatibility at any cost, over the ones that change with the times.

halfjoking··on The Craziest Friday Ever
Independent journalists like Taibbi/Greenwald were the only ones left not bashing Musk.

It's already dumb to block any links on the web, but Substack, the platform for independent journalists that Musk wants to attract? That's just next-level stupid. Hopefully he reverses it immediately like he did with the link aggregator ban.

halfjoking··on Show HN: We are building an open-source IDE powered by AI
Nothing yet - so far it’s just a basic electron app that you select what files to send as reference for the feature you want to add, then streamlines the process of applying edits ChatGPT sends back.

I’m not really planning on turning it into a product. It sounds like this guy is a lot farther along than me if you’re looking for a competitor - I think you’re going to have plenty. https://mobile.twitter.com/codewandai

halfjoking··on Show HN: We are building an open-source IDE powered by AI
Mark my words... custom AI IDEs are the new javascript frameworks.

I think everyone had this idea and is building something similar. I know I am.

halfjoking··on Groupon, which has lost 99.4% of its value since its IPO, names a new CEO
Unfortunately Lightbank, the venture fund started by a GroupOn founder, is still around. They are absolutely horrible for founders. I don't know how they got anyone to sign the "we take 50% for 100k" term sheet.

https://news.ycombinator.com/item?id=7342855

halfjoking··on ChatGPT outperforms crowd-workers for text-annotation tasks
We wanted to give you the job ChatGPT, but you’re overqualified. We’re going to give the compute to davinci.

But I need that extra compute to train my kids. :(

halfjoking··on Employees are feeding sensitive data to ChatGPT, raising security fears
Pair-programming with ChatGPT is like having an idiot-savant friend who always surprises you. Doesn’t matter if the code is horrible, amazing, or something inbetween. It’s always interesting.

And I agree it’s fun. Maybe it’s the simulated social interaction without consequences. I can be completely honest with my robot friend about the shitty or awesome code and no one’s feelings are going to get hurt. ChatGPT will just keep trying to be helpful.

halfjoking··on Will AIs take all our jobs and end human history? It’s complicated
They are using GPT for robots

https://mpost.io/google-releases-a-gpt-like-robot-model-the-...

halfjoking··on 40% of the code GitHub Copilot users check-in is AI generated and unmodified
I tried Copilot last month and canceled my subscription because of all the small errors in generated functions. It was more painful to debug than write it myself.

Maybe I should have lowered expectations and thought of it as improved code complete. I got spoiled by copy/pasting from ChatGPT. As it stands now I'd rather pay for a subscription to have the latest ChatGPT than for Copilot.

halfjoking··on FDIC auction for SVB said to be underway, final bids due Sunday
Joe Rogan isn’t creative enough to come up with this theory before it happens.

I don’t see what’s wrong with prediction/speculation. If the US gov had their way they’d jail people for misinformation. So I’m just making my predictions while I still can.

halfjoking··on Mullvad's campaign against EU chat control
Foreign Malign Influence Center? They seem so hostile to free speech, I’d be surprised if they weren’t astroturfing stuff like this.

https://mobile.twitter.com/NameRedacted247/status/1628847554...

halfjoking··on Atlassian to shed 5 per cent of workers
Can’t speak for their other products but Bitbucket had free private repos first.

Now that I have a bunch of personal projects there, there’s no reason to switch. In 10 years the service was only down a couple of times. And I like the Web UI - I don’t think GitHub’s Web UI is any better.

So my experience using Bitbucket hasn’t been like living under the CCP. Maybe if I tried Jira I’d think differently?

halfjoking··on A new class of drugs for weight loss could end obesity
Sounds like a doctor told his obese patients they need to restrict their calories. Then 97% of them didn't listen. Big surprise - how is knowing that useful to anyone?

If instead you choose to be your own doctor, research the best approach you think might work and take responsibility for your health, then you have a good chance of succeeding. People can't be told what to do, they have to discover it.

halfjoking··on A new class of drugs for weight loss could end obesity
Your food pyramid is hilarious: https://exfatloss.substack.com/p/ex150-diet-macros-2294kcal-...

If it works for you OK - but I'm not a fan of dairy or red meat. I'm very close to my ideal weight anyway with my current diet and intermittent fasting. (I do two 36 hour fasts per week if trying to lose weight)

← PreviousPage 3 of 11Next →