sieve

> P2P payments

Some P2P is between family/friends. Or you are paying rent etc. But a lot of it is business transactions that move funds from one savings account to another because the auto driver is accepting a payment in his mother's or wife's account or something similar. There is so much friction in the banking system if you have to create a current account that this is pretty common.

The system needs to improve drastically to support the two sides instead of letting it be. Should help your use case as well if that happens

Pennywise, pound foolish decision.

India is forced to subsidize farmers to the tune of $37B JUST for urea. Governments routinely offer free bus services to women, free cash handouts to women, free electricity to farmers (who then use the power to pump out groundwater and grow paddy in areas otherwise not suitable for it). The list goes on.

A $1B subsidy to eliminate friction on the payment front is peanuts.

I noticed this as well in the 2000s. Stuff like WinHex which I was pretty sure then was written using Delphi. Was into RE during those days, so I used to poke around the internals of software just for fun

There are pros and cons to both approaches. I moved from stack to register because I was writing the code gen backend from scratch in any case and preferred to have a register representation instead of converting from stack to register (which has to happen one way or another). Also mildly helpful with control flow analysis during bytecode generation and verification phases.

Nice! This is what I keep telling everyone: LLMs let you build stuff you may not have the time/energy to before. You still need to do the last 10% yourself though. Tying up all the loose ends.

Syntax is a personal choice. Can always be changed. Architectural choices are difficult to manage later on.

The VM seems to be a stack VM. I have my own python replacement project (https://news.ycombinator.com/item?id=48090665) that I am developing privately for now. LLM-assisted ofc. Started with a bytecode VM and then switched over to a register one. The biggest break from regular language systems was the decision to move to an Erlang-style preemptive scheduler.

I have read almost every kind of fiction under the sun. My standard for what is great is pretty high. This is why I am perfectly willing to accept LLM output as lot as the writer has put in sufficient work into the same. Because this kind of writing can produce middling to good enough fiction, like most genre fiction. Which I enjoy for the plot, characters and environment.

If a great writer used LLMs, then the output would be outstanding. Of that I have no doubt.

> The 'window' is I assume a french door, which I suspect is a mistranslation.

It is a french door/window. The original story is by Saki (HH Munro), one of the greatest short story writers of all time. He was English. The story is literally called "The Open Window."

> Sorry, I'm not doing all that work for you

Well, you asked. And I gave you stuff that I generated for comparative analysis. I am quite satisfied with the output. Most human writing, including genre writing, is subpar. LLMs are far superior to that. If you want to read only Faulkner and Eliot, then stick to them I guess?

> human writing examples

You must provide a solid prompt. And you must use a strong model. Deciding LLMs suck at writing based on output from Gemini 2 Flash, and expecting writing like Graham Greene is a bit much. The day LLMs start writing like him based on one-line prompts, writers will have no option but to look for other jobs.

Did you read the original by Saki (Exhibit 12)?

Ideally, you should read all 14 exhibits. Each has been produced by a different LLM, some local, others frontier models of the time.

E4 = Claude

E6 = ChatGPT

E8 = Mistral

E11 = DeepSeek V3

E12 = Original

This is something from early 2025. The situation and my own way of driving the models has improved somewhat in the intervening period.

Variations on a Theme of Saki

https://gist.github.com/s-i-e-v-e/b4d696bfb08488aeb893cce3a4...

My solution to this, as the human-in-the-loop, is pretty simple. If the generated draft is missing something that was implicit in the chapter outline, I ask the LLM to incorporate it and rewrite the para/section. Then I fix the outline to make the point explicit so that a reroll does not make the same mistake again

Yeah, if I think something is worth publishing, I might do that. I don't do this for the money. I do it because I enjoy building a completely lived in world from character sheets and outlines

> Producing a bunch of filler content to pad things out is the opposite direction from most of the reading that I have found helpful.

Padding, though it is probably the wrong word for what the LLM does when it expands the outline, lets you experience the world. The modern 8-episode tv season, for instance, might be technically well done but does not leave you with sufficient time in the world as compared to the 22-episode season with a bunch of filler episodes

Because I care about the world and the characters? And want to read stories set in that world instead of five pages of bullet points?

> But they generate objectively bad text.

They don't. I don't think people are taking full advantage of their capabilities. They produce perfectly functional prose. Sometimes, they produce something genuinely amazing to read. But that is quite rare.

People should stop expecting brilliant prose from LLMs without doing the preparatory work themselves first. Very few human authors can manage that in any case

> If what you care about is the characters and the plot, you already have those, you wrote them yourself.

Nah. What I have is the skeleton. You want it made flesh. That is what the prose does. It produces a fully realized world. I do not care to read stories in the form of bullet points. Which is what my character sheets and outlines are.

> There is _some_ stuff you read for the prose.

By this I mean you read it for the way the writer writes: prose qua prose. Most times, perfunctory prose that serves as the glue is sufficient

I do not expect the LLMs to do everything. They are assistants. You still bear the responsibility of stitching everything together into a cohesive whole. Make changes as required.

That is where expectations differ I guess.

There is some stuff you read for the prose. Most books, I read for the characters and the plot. So the LLM-assisted output works for me.

I tried the programmer's method. Automating stuff. Then I realized I prefer the wood carver's method. So I take it slow.

I think and write the outline for a chapter. Then I have the LLM produce it. I read the draft. Tweak things. Reroll. Take the best from multiple rolls and merge them to produce a organic whole. Manually patch the minor differences in pacing and tone.

Rinse and repeat.

It takes 100-250 hours across 3-4 months to produce a story. But I drop 99.9% of the stuff within the first day or two because the plot or characters do not work.

I use the web front end directly. Let's me switch between any model including Gemma 4 running locally.

> inconsistency across various parts of the story, items that are introduced, questions that go unanswered etc.

This is a workflow problem.

What the LLM normally needs:

- very basic outline: what the story is about

- basic cast of characters: like fiction/plays of the 18-1900s

Then you go chapter-by-chapter. But the most critical thing is that you MUST provide the chapter beats. From start to end. Location, environment, events, any critical conversations. For a chapter that is 2-3K words, this chapter outline often becomes 2-400 words or more. Because you have the story in your head, you can handle the stuff that the LLM is likely to screw up.

There are ways to build a system where the LLM can keep track of many things, but then that becomes programming, which is just as boring.

Think of it this way. You like well-designed houses. Like looking at them, living in them. So you design a house per your liking having bought a plot overlooking a stream and a meadow. And you hire someone to build it for you. You may not enjoy the actual process of building, but you do like living in the final product, right?

It is the same with books. Reading a story based on a plot of your design, and characters you built, dealing with situations you conceived of, and having conversations you directed works for me ... even if an LLM assisted with the writing duties.

I have read thousands of books in my lifetime. I don't really have too many hobbies outside of reading. Trust me when I say there is an audience for literal crap. LLMs actually raise the bar significantly. I do not think they will ever exceed Hugo, or TS Eliot, or Borges, but they are far superior to the random author publishing random crap every single day.

Can the writer use AI to improve their output ... is the only outstanding question.

> It's literally an art.

So was making clothes, or shoes, or cabinets, or knives, or cooking, or so many other things. Now we mass produce all of them.

You could say that for any book.

People are too precious about man-made art. I don't see the point. Every book ever written might have some novelty value of its own. But it mostly has the same plots, character archetypes and other tropes found in other fiction. We think there is something special about human writing. But LLMs have read the same stuff we have. And produce similar inspired writing.

I have hundreds of plot ideas from 40 years of reading and watching crappy books and films/tv. Never had the time or energy to explore those. LLMs let me try out various plots. To see which ones work and which don't. I haven't really thought about publishing anything though. Too much work and bureaucracy.

The thing is ... ideas matter just as much as the writing. You cannot simply say: "Hey Claude, write me a Sherlock Holmes pastiche set in the bronze age."

I have found that a 1:4 to 1:8 ratio gives nice results. That is, for every 4-8 paras of expected output, you must provide 1 para worth of character detail, plot beats etc. So you would need 100-200 pages of your own writing to produce 800 pages worth.

But most people are lazy. So the output is too.

I often produce have LLMs extract the text from a book scan and then translate it into English or some other language I know. Then I load it into my interlinear reader for side-by-side reading. Convenient when you are starting out with a language

Everyone should learn Sanskrit if only to read the great poets and dramatists in the original. The translations are often nice, but you do lose something in the process. Thanks to the incredible efforts of Pāṇini[1], the language was mostly frozen in time and is still just as understandable today as it was 2500 years ago.

I started learning it because it is India's civilizational language and I wanted to read the itihāsas in the original.

[1] https://en.wikipedia.org/wiki/P%C4%81%E1%B9%87ini

Melville's "Bartleby" is very good

> Can you name three titles that you would recommend?

I can recommend some authors from the period: W Somerset Maugham, Thomas Hardy, Anthony Trollope, H Rider Haggard, Wilkie Collins, Saki, Maupassant, O Henry.

Dickens, Hugo, Austen, Eliot, Bronte, Wilde, Doyle, Poe, Melville, Verne, Stevenson, Wells, Christie, Chesterton, Burroughs, Lovecraft, Chekhov, Dostoyevsky, Tolstoy etc are probably well-known.

The list could go on forever, almost. Some books might feel dated (or too long). But that is fine.

> how did you get into this?

I started reading when I was 3-4 and never stopped. We moved to the suburbs when I was 7-8 (late 80s) and there wasn't much else to do outside of school except read. Built a physical library of close to 4,000 books along the way, but I mostly read at the computer or on e-devices

I installed this the moment I got my Kobo Libra H2O. Mostly read fiction from the 1800s and early 1900s and I did not want to fight whatever proprietary software the reader came with (Kobo is probably better than the Kindle for this, but I never tried it!)

I have a problem with epubs though. This need to support every single thing under the sun makes the pipeline exceedingly complex. All you need for a vast majority of fiction is:

- headings

- paras

- emphasis (bold/italic)

- block quotes

Add simple "image" syntax to it to display maps etc from detective stories

Basically a ruthlessly cut down version of markdown.

The mini-readers from Xteink came in at the right time for such experimentation.

Indian companies do not like to do research. Too much risk. They prefer to buy proven stuff. There are small companies doing a lot of what the Chinese are doing (including LFP and other battery chemistry). But they cannot scale up because they lack the necessary investment and order book.

GOI has a new locally-produced solar cell mandate that has been pushed back by six months because the requirement is 8-10x the current production. People are being forced to keep factories shut due to the lack of locally produced cells.

https://www.reuters.com/business/energy/indias-solar-push-id...

It is a name. The identity comes from the rituals, beliefs, clothing, literature, temples, architecture, food and so many other cultural and civilizational markers.

Pakistanis who were looking for their fathers in Arabia, Persia and Turkey in the past based upon a deliberate avoidance of their Indic origin have recently started a political project to reclaim their identity based upon the IVC. They can try all they want but civilizational continuity is the only thing that matters. Names are dime a dozen. If you have a problem with Hindu/India and other exonyms, you could always call yourself Bhāratīya or some such. It is all the same.