HNHacker News
TopNewBestAskShowJobs

losvedir

10,735 karma · joined June 8, 2009

Senior Software Engineer at PagerDuty, using primarily Elixir and JS.

Previously senior developer at the MBTA, and before that co-founder and Lead Engineer at CoachUp, a Ruby on Rails start-up in Cambridge, MA (TechStars Boston F12).

Feel free to shoot me an email about anything: gabe@durazo.us.

The postings on this site are my own and do not represent PagerDuty's positions, policies, strategies, or opinions.

submissionscomments
losvedir··on Claude Haiku 5.5
Nah, it's pretty trivial to switch providers (especially with Claude's help, ha).

This is more to encourage people to try out adding AI into their product, which is a totally different flow and experience from using AI to build the product.

losvedir··on The lamps in my house
That was exactly my intended point (and my comment does mention the hour subdivision). There are a lot of meaningless coincidental homonyms, but these tend to not be homonyms in other languages. E.g. I dunno "left" in English would be "izquierda" or "dejó". Or conversely, "vino" in Spanish would be "wine" or "came". The fact that "second" and "segundo" have homonyms in their respective languages is a clue that the unit of time and the ordinal word are the same is not a coincidence. And when you look into it it's for the reason that you state. The homonym pairs indicates an etymological relationship between the two concepts.

So I was wondering if the same applied here. My initial reaction is that "light" meaning "illuminated" and "light" meaning "not heavy" is that this is just one of those coincidences, since I can't see how that one word relates the two concepts. But given that Japanese also does the same, it makes me wonder if maybe there is some sort of relationship. Maybe it was a common belief historically that the sun was not very heavy or something like that. Or maybe not and it's just a rare double coincidence!

losvedir··on The technology to eradicate mosquito-borne disease exists
Are you proposing that it would be infinitely bad? That somehow extincting a few varieties of mosquitoes could cause the collapse of all life on earth?

I don't get the "but what if it's somewhat - not infinitely - bad" argument? The known good is so dramatically good that we'd need a shocking amount of unknown bad to have made it not worth it. But if I'm choosing between mosquitoes, and say... fish, frogs, and a variety of birds, yes, get rid of the mosquitoes.

losvedir··on The lamps in my house
Unrelated aside, but:

> Fun fact: “Akari” means light in Japanese, both as illumination and as lightness.

Wow! I love it when two languages have matching homonyms. I always wonder "what does it mean ?". Like, in English "second" to mean the unit of time and the place after "first" feels like two distinct usages. But it's the same in Spanish ("segundo"), which points to a deeper connection (and isn't so unsurprising given the shared roots of the languages): the unit of time is called that because it's the second subdivision of an hour.

The two usages of "light" in English are so distinct I'm quite surprised it's like this in the entirely unrelated Japanese language. Is this just a big coincidence, or does anyone know any deeper reasons here?

losvedir··on The Legend of von Neumann (1973) [pdf]
von Neumann is a warning for me related to AI. Here he was, the smartest man alive, and he wanted to preemptively nuke the USSR. Thank goodness "less intelligent" heads prevailed.
losvedir··on Gemini 4 Argon (High): Intelligence, Performance and Price Analysis
The only models beating it are on "max", while this is "high". There's no guarantee that those effort / reasoning levels compare, but there's almost certainly an "xhigh" or "max" version later which will score higher there.
losvedir··on Gemini 4 Argon
> expanding the model’s output token limit to an industry-leading 1M tokens, up from the previous 64K tokens

Can someone help me understand this? I might have an out of date mental model of how these things work.

Fundamentally, LLMs output tokens 1 at a time, generating the next token from all the previous. And as the context window gets larger, this gets harder / slower / more expensive. So I get the idea of a maximum context window.

But I don't understand the point or meaning of an output token limit. I thought it was more a measure of price capping (since output tokens are more expensive) that a user could configure. I guess a model will keep generating tokens until it hits a "stop", so does this mean it's tuned to more aggressively produce output tokens? How does that fit into agentic loops. Are output token limits based on how long until it goes back to the user? Or does each "turn" of tool call, thought, tool call, thought, etc, get its own limit?

losvedir··on Gemini 4 Argon
They do release the model weights for Gemma. But could anyone actually run Gemini 4 weights? It's probably like a 10T model, which you need an industrial rack for anyway.
losvedir··on Ask HN: What are you reading?
Book 2 of Hyperion Cantos. Really enjoying it, but am unsure if I want to do books 3 & 4. Anyone read them and recommend either way?
losvedir··on Sonnet 5.5
Exactly. I'm saying that Sonnet 5.5 might not be useful or necessary in a Claude Code session but it could be good value in the API when you pay per token.
losvedir··on Sonnet 5.5
Useful for API requests, when using AI in the product rather than to build the product.
losvedir··on Does Reddit have an astroturfing problem? What the data suggests
What's their business model? A priori, I don't trust them because there's no reason for them to give out all that information for free. I assume it's a affiliate marketing / referral revenue stream? I suppose it could be fair within those parameters, but it would mean there's no incentive for them to test products that don't have such a program.

I pay for Consumer Reports, since you actually need a paid subscription to access the data, which makes sense structurally. I don't really know if the reviews are any good, though I've been relatively happy with it. I also paid a subscription for an independent supplement testing lab, but I forget what it's called at the moment.

losvedir··on Plan mode is dead
> In Claude Code, all plan mode does is add a little reminder to every user message

I was quite surprised when I learned this (when Claude Code edited a file despite being in plan mode). I had previously assumed "plan mode" was a harness level concern, and restricted what tools could be used. I didn't expect that it was simply an addendum to the prompt.

But I think it gave me some good insight into where the heads are at of Anthropic employees building this. Basically, leaning on pushing everything to the model. That's why alignment is so important: a "sufficiently advanced" model doesn't require any tooling infrastructure around it, and I suppose Claude Code devs are targeting that future. I had previously thought there was more to a harness, but with "auto" mode these days, it seems like there's no desire to build in that direction.

losvedir··on Claude Opus 5.5 Intelligence, Performance and Price Analysis (Max)
This index doesn't have "Astra" and "Opus 5". Every entry with corresponding data is a `(model, reasoning)` tuple.

So I'm unclear what you're actually saying and wondering if you've missed that. Are you saying that at every reasoning level it says Opus 5 beats Astra? I just compared Opus 5 high to Astra high and it has Astra as generally better than Opus.

losvedir··on Why I'm still bearish on LLMs after Navier-Stokes
> though it takes longer to devolve into illegal moves

Is this because the context is being saturated? How did you set it up?

Was the prompt something like "Here's the state of the board, you're white, your move, what do you do?" and then starting fresh each time? Or did it include the whole history of moves and board states and previous thinking tokens and so on? No judgment, just trying to add this data point (thanks for sharing!) to my mental model and understanding.

I'd be curious how it would work if it started fresh each time. My guess is it would never make an illegal move, although it may not actually play all that well.

losvedir··on Things That Annoy Me About Cars
I don't get the allure of one pedal driving. I have a Model Y, so I use it, but I'm just kind of "meh" about it. I think they introduced it related to re-gen, but these days it's kind of an arbitrary UX choice - on the Tesla you can enable brake assist to mimic the one-pedal driving experience when it's cold or you have a 100% battery, and other cars (maybe Tesla?) will periodically use brakes instead of regen to dry them off or otherwise exercise them a bit.

So then, if it's not actually a "fly-by-wire" mechanical system, why would you choose that as the UX? It cramps my calf constantly keeping it partially activated and the pedal half way depressed. More often than not I'll relieve my foot by doing short stints of cruise control, but that's tedious to manage.

My dream set up would be the accelerator pedal is, well, an accelerator, and the brake is a decelerator, and if you're not pressing either the car maintains the current velocity. It feels like _most_ of the time that you're driving you're just trying to make it go the same speed, and then you can press whichever pedal you need if you're wanting to stop or start again.

I don't suppose any cars work like this? ICE vehicles more or less in the moment do but "coasting" does decelerate a small amount so that over time you have to push the pedal again.

losvedir··on What do Visa and Mastercard do? An intro to card networks
You say "Correct", but Toast works the opposite of the person you're responding to. They charge the same fee regardless if it's a debit or credit card. They charge based on the network, and most debit cards in my experience use the Visa or Mastercard network. Yes, this is contrary to the Visa/MC terms, but I've never seen it enforced.

https://support.toasttab.com/en/article/Are-customers-charge...

losvedir··on What do Visa and Mastercard do? An intro to card networks
Debit cards aren't that, in practice, because as you say they typically get the same 3% fee whenever I've asked. Toast and other point of sale networks just charge it based on the network and not the card type.

But even so, I still prefer the credit card instrument, used as a "charge card". For those that don't know, a "charge card" is common with businesses and is expected to be paid off every month. But it helps with managing your cashflow - e.g. you can keep your money in a savings account all month and make one transaction at the end - and it keeps your actual money from being at risk of fraud.

Most Europeans seem confused with how Americans use credit cards for everything, but about half of us use them as charge cards, paying them off every month.

That is a benefit, so I can see paying a small percentage for it, but I don't think that benefit alone justifies the 3-5% of a transaction that credit cards charge now.

losvedir··on What do Visa and Mastercard do? An intro to card networks
I play the credit card "game" because I hate feeling like I'm leaving money on the table, but I really wish I didn't have to. Some degrees of fees make sense, to run the network, and handle fraud, but it's ridiculous that people have this sense that shopping should somehow fund your vacation.

Where I am it's increasingly common to see credit card fees when checking out. I get it, because merchants are being charged 3-5% of their total revenue. I wish there was a low fee credit card network that merchants didn't charge a fee for, so I could continue the simplicity of digital payments but opt out of this crazy Visa Infinite rewards accounting boondoggle.

losvedir··on Is mathematics about to enter the conservatory?
This is confidently presented, but seems incorrect based on other commenters.

> Snark noted, though I have no earthly idea why you felt it necessary

I think it's useful if you are, in fact, talking out your ass. It's hard to tell given your confident phrasing, but saying something which gives away that you don't actually have meaningful experience in the subject is worth calling out. It sounds like your interlocutor has a better sense of the state of the music world, and that was a way for them to indicate it.

losvedir··on Thoughts on Pope Leo XIV's Magnifica Humanitas
The abuse, sure, but the birth control stance is really the only one compatible with the Catholic idea of sex only being appropriate in the context of a monogamous, lifelong married couple. I'm not really sure what you're getting at with that one, since there wouldn't be an HIV or STI epidemic in that world.
losvedir··on MIT's Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training
I think you can tell who still has the ability to read long form content by whether they thought this was full of fluff or decently actionable. I think the bolded parts of section 3 are definitely not fluff. I found the document pretty interesting.
losvedir··on Tim Curry has died
Everyone's got their Tim Curry role that defines him. For me it's from The Three Musketeers. Though Nigel Thornberry is a close second.

You can probably identify the age of every commenter in this thread based on the particular role that sits with them.

losvedir··on Dolly Parton has died
RIP. We went to Dollywood last year and I highly recommend it! It was a fun, family-friendly park, and one of the things I was impressed by was how helpful and happy to be there all the workers were. Seemed like a bunch of friendly veterans who took our tickets and let us in.
losvedir··on How credit card rewards became a $9.2B wealth transfer
I think that's out of date. He links to a study showing interchange revenue net of rewards showing up to 3% by high FICO scores. (Just at a gut check that seems crazy to me, since interchange revenue doesn't really go much above 3%!). But that's from 2013. I remember when Fidelity launched its 2% flat cashback AmEx back in 2003. People didn't really know if it would be sustainable. Now 2% is a dime a dozen.

The most recent I've seen otherwise is this Federal Reserve study[0] from 2022. It finds that the marginal return on swipes is actually slightly negative because of how juicy rewards have gotten, and 80% of their profitability comes from interest (with most of the rest fees):

> we find that, on average, the credit function makes up approximately 80 percent of the credit card profitability, whereas the contribution of the transaction function is slightly negative, as rewards and other expenses on credit card transactions outpace banks' interchange revenues.5 In addition, fees—in particular late fees—comprise approximately 15 percent of credit card profitability.

[0] https://www.federalreserve.gov/econres/notes/feds-notes/cred...

losvedir··on How credit card rewards became a $9.2B wealth transfer
The rewards differences can be significant. Eg 4% vs 1% cash back is $3k difference on an annual spend of $100k.

On the one hand, relative to our income it's not so important, but on the other it feels bad leaving $3k on the table.

losvedir··on Scrap (2006)
Heh, $600k still sounds like a lot to me! Talk about a bubble.
losvedir··on Pixel 11 Pro Fold feels like the end of an era
The point of decoupling the phone from the contract is you don't buy your phone through your carrier. Why would you want to do that?
losvedir··on Pixel 11 Pro Fold feels like the end of an era
I can't tell if you've actually tried a finding phone or not, but that would help understand and place your comments better.
losvedir··on Opus 5.0 drives incoherence into the stratosphere
Same, oh God the leaking of our chat infuriates me. It also loves to document public functions with what the caller is and doing, and I'm like that's not how this works! The point of an abstraction or interface is that it stands on its own. It's not supposed to be for some specific caller elsewhere in the code.
← PreviousPage 2 of 34Next →