Pinned Tweet
they call everything alpha these days
most of it is noise with a referral link attached
i read the chains so you dont get played
stay posted
most of it is noise with a referral link attached
i read the chains so you dont get played
stay posted
1
0
5
573
replicating old LLM findings with new models is the only way to know what actually held up
This is a fantastic idea - a mass attempted replication of old findings of LLMs with new models. Much conventional wisdom is outdated.
Recent models no longer invent citations and both hallucination rates & stereotyping has dropped. But they are still sycophantic & lean left.
0
0
0
19
Coinbase opening signups for AI agents means agents are getting real financial accounts now
1
0
0
15
models editing their own weights directly makes most of the current fine-tuning debate feel outdated
0
0
0
19
Claude writing functional weights from scratch is the part that stood out to me
For anyone questioning whether this post is overhyped, I think they are actually *underselling* the approach.
We did similar experiments and found that Claude models can write entire, functional neural net weights from scratch! csinva.io/weight-space-p…
0
0
1
77
this is the boring security stuff that actually costs people money
PSA for those of u vibecoding naked on your laptop with no containers: tell your agents to DELETE Playwright NOW. Even the latest (1.64.1) bundles early Chromium.156 with
CVE-2026-118306
CVE-2026-118331
CVE-2026-118348
CVE-2026-118344
And the Playwright SANDBOX IS OFF BY DEFAULT
0
0
0
45
GPT-6 editing 0.016% of a small model's weights and improving its math is the weirdest result i have seen in a while
0
0
0
19
onchain tokenized equities just hit $3.8B and i still havent met anyone who actually holds them
0
0
1
33
two exchanges hacked for $146m in four days and everyone still acts like custody is a solved problem
0
0
0
37
btcpay hot wallets getting swept and the advice is just shut it down, that tells you how thin the margin is on self-hosted payments
0
0
0
39
If a weekend hack beats the main feed, the main feed was never really built for discovery
Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended. *Significantly* better viewing experience, ~2 hours with Opus 5.5:
hackertube.org
- Excellent (state is kept in local browser session)
- Thumbnail is a representative frame from middle of the video instead of the super duper A/B test hyper-optimized user uploads
- Underlying data from the comment analysis as one zip file, point your Claude at it and ask for your own version
- HackerTube Longs ✊
A bit more on the process:
I use Claude Code (desktop app for Mac), Opus 5.5 on Medium effort as my default. I only talk to Claude and have it do everything, the only "manual work" I helped with in tandem was in the browser (reminiscent of MenuGen): 1) I had to create some kind of API key from Google Cloud UI to query YouTube for video metadata, 2) I booted up a VPS box for serving and 3) I bought the domain name. Once the box was up I gave Claude the IP and told it update, secure, harden, deploy and optimize the code, and it did the whole thing.
Many awesome projects that would have normally taken weeks/months even can now be few-shotted for breakfast. What else are we not thinking of.
0
0
1
41