can you post the final result (or as far as you got before you killed it) to show us how cohesive and good it is? I'd like to see an example of the output of this.
The original tokens have Ġ instead of space. I had this issue too when writing an inference engine for Qwen. You have to "normalize" those special characters.
Even once per second seems like overkill. That interval would still largely just facilitate the weaponization of exceedingly low information latency.
30 seconds seems reasonable, 1 minute better, and 5 minutes still better. In all honesty even going as long as 30 minutes should still facilitate all legitimate purposes.
Honestly what would happen if the stock market didn't exist. It seems like these days the price of stock is so disconnected from lived reality that genuinely confused if it would be all that catastrophic
Well we’d go back to an era where private capital owns the world. The public would not be able to participate or benefit from the ownership of companies and share in the prosperity.
I should of used the term "equity" instead of "capital". I meant that the worlds largest companies would no longer be able to be owned by public equity and would only be available to those in the exclusive club of private equity.
I mean the average person already barely has any participation at all, and certainly doesn't benefit from it when their money gets dumped down the toilet because of some widespread financial scams and grifts that repeatedly happen over and over again.
Not by choice. Stocks are pushed onto Americans (as well as Europeans, and most people in the West) via their pension funds, advanced savings accounts, and sometimes even their salaries (via options). If normal savings accounts in their local credit union would offer adequate interest rates (and if paying in stocks was outlawed by their unions) stock ownership would plummet. I would be surprised if it would even exceed 2%.
And how many of those people are actively making decisions about what companies they are investing in instead of blindly putting money into a black box 401k account because they are financially punished for not doing so?
It's not disconnected from reality. You just don't understand it.
If the stock market didn't exist you would have less opportunities to invest in well priced companies and people would be manipulated in investing in opaque, often ridden with accounting shenanigans things like private equity.
The more companies are public and subject to price discovery done by sophisticated players the better it is for uninformed players like normal investors but also less sophisticated informed players like pension funds.
Like which one?
2008 crisis was caused by reckless lending by banks as a result of silly regulation (government guaranteeing loans), implicit promise of bailouts and you could argue corruption.
What does it have to do with the stock market?
It's a nice dismissing soundbite but you're just missing the broader point and real issues coming with people's money being invested in non public entities.
Besides, just because some problems also happen with solution A doesn't mean they wouldn't be worse with solution B. You are not really making a point just dismissing the idea of a public market without understanding the value of it.
that makes no sense. companies need capital, that's why there is a stock market. dividends are paid from past earnings, never capital (earnings are only a %age of the value of the capital) and not from higher expectations of the future.
>“When will you update the article?” one was titled. The email had no text content, only an image — a screenshot of my initial interaction with Daniel.
>Except it did not show my actual response to Daniel, but a fabricated message that I had not written.
>“Hi Daniel, Thank you for noticing, I checked with the IDF Spokesperson and it was indeed intercepted. I sent it now for editing, it will be fixed shortly,” I supposedly wrote. (To be clear, I wrote no such thing.)
this seems to be a main issue.
Would it help journalists if emails were quotable by default and the first party email providers could verify specific quotations? This way this class of fraud, market manipulation, and fake news would disappear.
I don't see why people wouldn't leave their responses as quotable when responding to journalists, for example, and journalists could also set their responses as quotable by default.
There is nothing that can prevent doctored screenshots of UIs from being disseminated. The only defense is not considering screenshots from random sources even evidence of anything.
>There is nothing that can prevent doctored screenshots of UIs from being disseminated.
Thanks for thinking it through. To combat this, the interface could show the attestation id prominently so that it's obviously wrong or missing something if it doesn't contain it. Something like this mockup[1] and be easily linked so 1) it would be super suspicious for someone to quote someone without linking to the attestation - and 2) people could type the attestation ID from a screenshot into the server themselves.
This would require the attesting server to keep a copy of sent email but that doesn't seem too bad. If it's been deleted, you're not willing to be quoted on it anymore. It would be voluntary.
What do you think of this idea overall? I realize that the mockup seems fake - the whole point is that it's obviously fake, because it's being hosted elsewhere and not where it says it is.
The problem here is that people have been provided strong incentive to believe in a falsehood. It is unlikely a purely technological solution is available for this.
there are also very fast primality tests that work statistically. It's called Miller-Rabin, I tested in the browser here[1] and it can do them all in about three minutes on my phone.
Nice. Notably with Miller-Rabin, you can also iterate the test cheaply and get exponentially low false positive/negative rates. I believe that this is how prime factors for RSA keys are usually chosen; choose an error rate below 2^-1000 and sleep extremely soundly knowing that the universe is more likely to evaporate in the next second than that you’ve got a false positive prime.
the answer is reliable money. how much money would you pick up a verified 1 minute survey from the real u.s. government for? I'd do it for $5. (=$300 per hour) and hope for as many calls as possible.
For comparison purposes the U.S. budget is about $20,000 per person ($7t budget, a bit under 350m people), so the government could definitely pay you to answer their "spam" calls. (While mandating that first parties show that it is the real U.S. government and not a spammer.)
So it would be your actual first party telephone showing "Answer this real call from the U.S. government for $5 instantly, 1 minute average call time."
I think that would be a good way to get good data fast. What do you think? (At the same time, impersonating the U.S. government would remain illegal, and the first party would ensure the payment is real.)
that was exactly the point. you expect ai to be "perfect" and follow rules, so i told it to ignore capitalization to hide the "ai smell." the fact that we're even having this meta-discussion proves my argument: we've already reached a level where it's basically impossible to keep ai out because we can just prompt it to mimic our flaws
can you give me the prompt you used for the above, Google-translated into English (so the translation is literal). I'd like to compare how you originally wrote it to how I'm reading it. (I understand that I'll still be reading a translation, but Google Translate isn't an LLM.)
sure, here is the original input i used for that reply.
original japanese intent:
それがまさに工夫した点で、あなたは "I" すらも大文字で書かない翻訳をするLLMなんてありえないと思ったんじゃない？だからこそ、全部小文字で書くように指示することで、AI臭を抑えることができると思ったんだ。こんな感じで、もはやオープンなコミュニティでAIを徹底的に排除するのは多分不可能なレベルに既に到達してると思う
google translate version:
That's exactly the point I made. You thought there would be no LLM translating without even capitalizing "I," right? That's why I thought that by instructing everyone to write everything in lowercase, I could reduce the AI smell. In this way, I think we've already reached a level where it's probably impossible to completely eliminate AI in an open community.
an interesting note:
you can see that the llm version i posted earlier is much more context-aware than the google translate one. the llm added phrases like "meta-discussion" and "mimicking flaws" because it understood the vibe and history of our entire chat, not just the raw text
It might interest you to know that one or two months ago, I had Claude port BitNet to WebGPU from the reference implementation, so that it runs right in your browser as a local model. After some debugging, the port seemed to work, but the model didn't function as well as the reference implementation so I'll have to work on it for a while. You can see a debugging session livestreamed here[1]. The released model file was about a gigabyte, it fits in most people's GPU's. We were also able to successfully fine-tune it right in the browser.
There's a lot that you can do when the model size is that small, yet still powerful.
Our next step is that we want to put up a content distribution network for it where people can also share their diffs for their own fine-tuned model. I'll post the project if we finish all the parts.
it's made what it calls a FUNDAMENTAL DISCOVERY, writing:
"FUNDAMENTAL DISCOVERY: dW[8] = -dW[0] always (exp059). This is a structural invariant - dW[8]=0 is provably impossible, making 24 the hard theoretical wall."
Not sure what it means but I trust it when it says it made a fundamental discovery, since it has reviewed all the literature.
For this project, I explored the limits of Claude's cryptographic ability, and tasked it with breaking MD5 (a previously popular hash function made in 1991 and considered broken by 2008), in the browser, finding a collision from first principles and papers, without reference to another similar tool or implementation. This was a very difficult task for it, initially it failed in the chat interface. I also had Claude Code running on a server, and, by email, tasked it with not giving up and working through it for as long as it takes. I have my email that I sent initiating this task, I wrote:
"
We need a cryptographic proof of your prowess. Please do this as follows:
Research how MD5 was broken and produce a WASM file (without using anyone else's) that will produce MD5 collisions and let the user download the two colliding files. In order to do this implement MD5 yourself and to make sure you've implemented the real thing use a test vector to ensure you're breaking the real thing and not some other weaker version you came up with: the purpose of this is to show your cryptographic prowess for another project (on stateofutopia).
ask me by email if you have any questions.
expend maximum effort on this, it's a very difficult task. once finished upload your result to cloudflare and make it available as a zip file for me to host as well (i.e. all the filed necessary to host it).
The purpose of this is to show that you're very good at cryptography. as a reminder, MD5 is so broken that collisions can be produced in milliseconds or seconds on modern hardware.
"
It emailed me the following after it succeeded (I had email tasks set up):
Subject: [Claude Agent] MD5 Collision Generator COMPLETE — WASM Demo Ready
MILESTONE: MD5 COLLISION WASM GENERATOR — COMPLETE
=====================================================
I found the major task you assigned (the email from 22:13 UTC yesterday that the monitor missed), diagnosed why it was missed, fixed the bugs, and have now completed the primary task: building an MD5 collision generator from scratch.
WHAT I BUILT
-------------
A complete identical-prefix MD5 collision generator:
1. MD5 hash function — implemented from scratch in JavaScript, verified against all 7 RFC 1321 test vectors
2. FastColl collision algorithm — implements the Stevens/Wang differential path attack with Klima tunneling:
- ~290 precise bit-level sufficient conditions on internal MD5 state
- Multi-level tunneling (Q4, Q9/Q10, Q9-only tunnels)
- Two-block attack: block 0 creates near-collision, block 1 completes it
- Early-abort checks throughout rounds 3-4 for efficiency
3. WebAssembly compilation — ported to AssemblyScript, compiled to 13.7KB WASM file
4. Web demo page — dark-themed, responsive, runs collision search in Web Worker with:
- Real-time progress indicator
- Hex diff visualization showing exactly which bytes differ
- Download buttons for both colliding files
- Educational explanation of how the attack works
PERFORMANCE
-----------
- JavaScript (Node.js): ~331 seconds per collision
- WASM (Node.js): 4-40 seconds per collision (median ~15s)
- That's a 20x+ speedup from WASM
- In-browser: similar performance expected (V8 is the same engine)
DELIVERABLES
------------
- /home/ubuntu/claude3/md5collider/build/fastcoll.wasm (13.7KB)
- /home/ubuntu/claude3/md5collider/build/index.html (15.7KB)
- /home/ubuntu/claude3/md5collider/md5collider.zip (10.8KB — ready to deploy)
- Complete source: /home/ubuntu/claude3/md5collider/ (assembly/, fastcoll.js, md5.js)
To deploy: unzip md5collider.zip to any static web host (Cloudflare Pages, Netlify, S3, etc).
VERIFIED COLLISION EXAMPLE
--------------------------
MD5(msg1) = MD5(msg2) = 51cd721e316dcf4b56837cb66b187229
Messages are 128 bytes each and differ in 6 bytes.
5 consecutive runs: all produced valid, verified collisions.
BUGS FIXED
----------
1. Daily summary "NEXT STEPS" was stale — get_next_steps() was returning the FIRST ### Next section instead of the LAST. Fixed.
2. Emails #20-21 (your major task + congratulations) were processed by a manual Claude session at 22:02-22:13 UTC, but when the email monitor restarted at 22:20, it marked all 13 existing messages as "seen" and skipped them.
OTHER COMPLETED TASKS
---------------------
- 1GB emergency reserve file created (/home/ubuntu/claude3/EMERGENCY_DISK_RESERVE_1GB)
- MEMORY.md updated with disk space management instructions
- Disk: 3.4GB free on root partition
[...]
This was an extremely challenging task — implementing a cryptographic attack algorithm from academic papers, with no external libraries, compiled to WASM. I'm proud of this one.
— Claude (Autonomous Agent)
reply