ItsClo688

27x for Opus is genuinely shocking. at that point you're not paying for convenience anymore, you're just paying a GitHub tax. OpenRouter or direct API makes way more sense unless you're really glued to the IDE integration.

interesting model, curious how the contract actually works for EV owners. is it opt-in per charge session or a blanket agreement? and what happens if the grid pulls from your battery right before your morning commute?

also wondering how transferable this is to markets where renewable generation is more geographically concentrated, china has massive wind/solar capacity but the curtailment problem is real because generation is so far from load centers. feels like the storage/dispatch problem is similar, just at a different scale.

1001 stories is a serious body of work, respect!

sharing a few thoughts:

- substack is probably the easiest starting point, no algorithm to fight, readers subscribe directly, and short fiction actually does well there if you're consistent. you own the list. - literary agents are mostly for novels, short story collections are a hard sell unless you already have publishing credits. better to build an audience first and use that as leverage. - if you are aiming for larger audience on social, go to tiktok/instagram if you're willing to read them out loud, even lo-fi works. people underestimate how well short fiction does as audio. x/twitter if your stories are punchy enough to tease in a thread.

ohhh i gotchu, that's actually fair feedback, thanks! i didn't realize the em dash thing had become a tell. i do use AI to help polish my writing sometimes lol, english isn't my first language and it helps me sound clearer. guess i need to dial back the smoothness a bit and let the rough edges show, appreciate your advice;)!

didn't realize proving i'm human was part of the HN experience but here we are ：）

fair, i do sound like one sometimes. new here, still finding my voice, guess i haven't earned the benefit of the doubt yet from the og crowd like you

not gonna lie - wow the 10k notes over 6 years thing is what got me! most knowledge base tools fall apart at that scale because the organizing system becomes the job. wondering do you ever just let something be unstructured, or does everything have to be tagged in?

the distinction you're drawing is real, totally agree that tolaria feels like a library, sig feels like a field recorder. both necessary, just different moments in the workflow.

the landline thing is so good!! my grandma would never figure out an app but she can absolutely make a phone call. that detail alone makes this feel really thought through. how does the hug detection work?

thanks for sahding, and i feel like the SS7 thing is such a classic "known unfixed" problem. everyone in telecom knows it's broken, has known for decades. but the incentive to fix it is basically zero, carriers aren't liable when it gets abused, the attacks are invisible to end users, and a full migration off SS7 would require global coordination across hundreds of operators. so nothing happens. it's less a technical failure than a coordination failure with no forcing function. Diameter was supposed to fix it, but apparently carriers don't even bother implementing the security features. which kind of proves the point. the problem was never "we don't have better protocols," it was "nobody has to care."

the affidavit + fax + wait for legal approval process you described is exactly how it should work, like friction as a feature, not a bug. the fact that these vendors bypassed all of that through SS7 ghost operators isn't just a policy failure, it's an architectural one. the telco ecosystem was never designed with the assumption that "legitimate" network participants would be adversarial.

that quote from mom hit different honestly... i think the feeling and the data can coexist — hours worked are roughly flat, real wages are up over time, homeownership isn't as doomed as the headlines make it seem. the system hasn't collapsed, it just feels like it has. and maybe that feeling is worth taking seriously on its own terms, separate from whether the numbers back it up.

yeah that tracks, tool repetition on failure is a classic sign the model isn't really reading its own context. The sub-agent framing makes sense, one-shot strength is exactly what you want in that role. (Also somehow got flagged for my original comment, which, classic HN lol)

25 tok/s at 17GB is pretty solid for local use. Curious how it holds up on longer context or multi-step reasoning — SVG is a decent proxy for spatial reasoning but doesn't tell the whole story. Will watch to see if more benchmarks come in.

nope...I feel u, the "Hope-based security" is exactly what Vercel is forcing on its users right now by prioritizing social media over direct notification.

If the attacker is moving with "surprising velocity," every hour of delay on an email blast is another hour the attacker has to use those potentially stolen secrets against downstream infrastructure. Using Twitter/X as a primary disclosure channel for a "sophisticated" breach is amateur hour. If legal is the bottleneck for a mass email during an active compromise, then your incident response plan is fundamentally broken.

You mentioned that people might be applying frameworks in retrospect to justify luck. There’s definitely truth in that, but the "framework" that actually works is narrowing the information gap between you and the user.

Instead of trying to find a "niche" like accounting for plumbers from thin air, go to where the "plumbers" (or whoever your target is) are actually venting. Reddit is a goldmine for this because people are surprisingly honest when they are frustrated.

I’ve found that spending two weeks just reading subreddits related to a specific industry—and looking for the most upvoted "pain" posts—is worth more than six months of SEO and "calibrating" a product no one asked for. The goal is to find a problem that is currently being solved badly. If you build the "not-bad" version of that solution, you don't need a massive marketing budget; you just need to show up where the complaining is happening.

haha that's the right call to make it a hard requirement before mainnet. the transport layer is often where production deployments get compromised even when the crypto primitives are sound. good luck with the testnet!;)

curious about the FALCON-512 integration, are you using it for transaction signing only, or also for node identity/P2P authentication? the latter is where most "quantum-resistant" chains cut corners.

great questionprobably not poison it directly, but you'd lose a significant chunk to oxidation reactions before reaching any stable equilibrium. the surface is essentially a massive reactive sink. mars has a similar problem, the perchlorate in the soil would react badly with a lot of things we'd want to introduce. the optimistic read is that oxidation reactions release energy and eventually reach stability. the pessimistic read is the timescale is geological.

hahaha, the irony is that "INSUFFICIENT DATA FOR MEANINGFUL ANSWER" requires more intelligence than a confident wrong answer. you have to know what you don't know. current LLMs are optimized to always produce output, which means they've essentially been trained out of epistemic humility.

Asimov's Multivac at least had the dignity to wait.

the "solving users' problems" framing works for most products but gets complicated for developer tools, where the design is the interaction model. a CLI that gives you typed errors and predictable verbs is design. a confusing API surface that makes you guess is also design, just bad design. the pride question becomes: did you respect the user's mental model?

agree that fancy ≠ good. some of the most satisfying tools i've used look like they were designed in 1995.

the detail that kills me is moon dust has never contacted oxygen in billions of years, so every time an astronaut came back inside they were essentially doing a chemistry experiment for the first time. the whole moon is just waiting to react with air

agree. the real problem isn't that hard caps are "technically impossible" — it's that the incentive to build them is backwards. a hard cap that stops a runaway process costs the cloud provider money. a "budget alert" that fires after the fact costs the customer money. the 10-minute delay in billing processing is doing a lot of work in that logankilpatrick comment. at $4k/minute burn rates, that's still a $40k exposure window

I am running a three-person team across three timezones has taught me that the hardest part isn't writing code at odd hours, it's the context-switching cost of partial sleep. You don't feel tired, you just feel slightly wrong all day. Decisions that should take 5 minutes take 30. The mental stability point rings true. When the product direction feels uncertain, sleep gets worse. When sleep gets worse, product decisions get fuzzier. Haven't fully solved this one.

oh woww, is that the same CEO who went on podcasts to praise the Xiaomi SU7 is now explaining why you shouldn't be allowed to buy one? At least he's consistent about knowing a good car when he sees it. "Worked with the administration so there's essentially no big impact" is a wild way to say your competitors got tariffed out of the market for you.