wxw

8 ms·

1.
by wxw 6d ago
> As part of the deal, Google signed a 22-year contract with Finnish utility Fortum to buy up to 50% of the output from the Loviisa nuclear power plant. > Finland has emerged as an attractive location for data centres because of its c
2.
by wxw 8d ago
Playing around with it tonight and I feel like a killer feature is its speed. Really fast response times.
3.
by wxw 8d ago
> The Early Access Network is an invite-only group of senior technology leaders who want to see what's coming before it's on anyone's shortlist. > Four times a year, we put a hand-picked group of YC's enterprise-AI
4.
by wxw 14d ago
“contributor” pricing at $0.10/$0.20 is crazy cheap if it’s measuring up to Sol. Definitely shows how important a user data flywheel is for RL and model improvement.
5.
by wxw 14d ago
> The rolodexes came from the now dispersed collection of former lab employee Ed Grothus, who operated a salvage company of lab cast-offs known as The Black Hole. Too good.
6.
by wxw 15d ago
> the True Rate of Unemployment tracks the percentage of the U.S. labor force that does not have a full-time job (35+ hours a week) but wants one, has no job, or does not earn a living wage, conservatively pegged at $26,000 (in 2025 doll
7.
by wxw 16d ago
How well does it work in practice?
8.
by wxw 20d ago
100% agreed. Small, cheap, and hosted models. Luna (and open weight models and others) is ridiculously cheap @ $0.2/$1.2, easily accessible, and more than good enough for basic use cases (e.g. summarization, simple tool calling, etc.).
9.
by wxw 21d ago
I just set up Tailscale as I've been hosting more personal apps off a small Hetzner VM and wanted a really simple personal networking solution. Works great, would recommend.
10.
Zohran and the Short Link(iamwillwang.com)
333 points by wxw 21d ago|217 comments
11.
by wxw 27d ago
Really enjoyed how thick this essay was. > The slop-acalypse that so many of us fear, when people will stop looking at anything made by humans and only look at things made by machines—why hasn’t that happened yet? Unfortunately, I'm
12.
by wxw 27d ago
> Although I’ve been saying “Claude”, I actually did most of the work in OpenCode with some of my favorite open weights models. Kimi K3 & K2.6, and DeepSeek v4 Pro & Flash. Interesting, is Claude becoming the Kleenex of the agent
13.
by wxw 28d ago
I think this is an interesting move in a world where Anthropic is leading the frontier. But I don't think we're in that world, at least anymore. The current vibe feels like OpenAI and Codex are leading the race. So this instead be
14.
by wxw 28d ago
> Notably, our worker did not have access to the web_search tool, but instead decided to use curl to access DuckDuckGo, Github, grep.app, and SourceGraph. Could this be fixed with better harness restrictions/tool sandboxing?
15.
by wxw 28d ago
> And as we grow, we will relentlessly aim to preserve the velocity, agility, efficiency, and talent density of the 90-person startup that we are today. Echoes of WhatsApp. Huge congrats to the team!
16.
by wxw 1mo ago
Perhaps has to do with the Github sync feature, https://cursor.com/docs/origin/mirror-github
17.
by wxw 1mo ago
These photos are beautiful: https://media.newyorker.com/photos/6a7f7ed55b6cfa7eff424505/...
18.
by wxw 1mo ago
> With its operating systems and its App Store, Apple controls a key infrastructure for the distribution of apps on its devices. In addition, Apple offers its own apps and advertising space. This dual role makes Apple subject to specific
19.
by wxw 1mo ago
I kinda miss the CSS bootstrap aesthetic.
20.
by wxw 1mo ago
> Scaling post-training is all we did for GLM-5.3. Love this opening line. And wow, great results. > As agent capability improves, much of the difficulty in scaling post-training moves from the model to the environment.
21.
by wxw 1mo ago
More detail in: https://x.com/XOpenSource/status/2087951962004230428 And the actual repo: https://github.com/xai-org/x-algorithm > We’re releasing the code that determines how posts are ra
22.
by wxw 1mo ago
> Compared with output speeds reported by Artificial Analysis GPT-5.6 Sol on Ultrafast mode runs 11x faster than Fable 5, and 5x faster than Opus 4.8 on Fast mode. Awesome work. I'm personally very excited for faster models/inf
23.
by wxw 1mo ago
Ah, multimodal is a great point. I'll need to try that some time.
24.
by wxw 1mo ago
They need to release benchmarks against Luna/Terra. Luna is much cheaper which feels like it undercuts the need for Flash. I've always considered the Flash series of models to be for low-cost, high-volume, mostly text-based use ca
25.
by wxw 1mo ago
> “McDonald’s secret sauce is really commercial surveillance,” says Jeff Chester Great line. I do really love their fries though.
26.
by wxw 1mo ago
> Previously, she was chief ethicist at Meta. She spent 6 years at Meta. I get the sense there are other factors at play than "ethics departments are just PR stunts". There's no way she isn't already aware of that. Un
27.
by wxw 1mo ago
Meta's clearly changing strategies back towards their original "frontier open source", but this time around they have a lot more competition from leading Chinese labs. I'm all for it though, and I think Glimmer is a fant
28.
by wxw 1mo ago
The top 3 names being > Freedom > Liberty > Serenity Reads like a state motto
29.
by wxw 1mo ago
> What you get is a beautiful animation that is 100% accurate and free of hallucinations. I'm not sure I follow how this is actually guaranteed? The fact-checking process mentioned just seems to involve asking AI to review its own w
30.
by wxw 1mo ago
> More broadly I’ve observed that as you grow older and gain experience in any domain, you’re rewarded for staying in your circle of competence and deepening the gyre. You don’t charm anyone by constant pivots, lest you seem unserious, u

More ›