dannyw

7 ms·

91.
by dannyw 2mo ago
It probably made its way onto a Jira backlog somewhere, and some engineer thankfully took it as a 1 story point task.
92.
by dannyw 2mo ago
All more reason to stick with standard TLDs like .com, .org, .net, and your country’s TLD when you can.
93.
by dannyw 2mo ago
I wonder how much of it is fears over "mass-hacking the internet", and how much of it is fears over the model discovering various NSA/CIA/etc "tailored access operations", and other deliberate side-channels &#x
94.
by dannyw 2mo ago
Total costs sure, electricity only costs no. My two DGX Sparks run DS4 Flash at about 50tok/s concurrency=1 which is more than suitable; at about 150W total wall power when generating. That’s about A$16 a month in electricity if I ran
95.
by dannyw 2mo ago
I agree they’ve done a bit too many pricing / usage promotions and A/B tests. The period was also marked with many billing bugs, like spending people’s usage credits for included Fable for a few hours (gave me a huge shock), but t
96.
by dannyw 2mo ago
Some models have been in long-horizon, multi-day LLM gyms for a long time, and across millions of sessions (if not in the tens, or hundreds of millions now). Some models have not. The former will perform well in these long horizon benchmark
97.
by dannyw 2mo ago
Nothing beats the classic of figuring out something yourself with your brain, but I also like dictating to LLMs a stream of consciousness with what I'm interested in (while forcing it to NOT give any answers or opinions), and getting b
98.
by dannyw 2mo ago
You've missed a really great human-authored piece then.
99.
by dannyw 2mo ago
11labs does _way_ more than basic TTS. The options, customizability, quality of voice cloning (for editing your own voices), etc is far more than what you see on the surface. Check out their docs for a glimpse: https://elevenlabs
100.
by dannyw 2mo ago
Set output style to explanatory in Claude Code. It's much better. I personally think it should be the default, but I assume Anthropic has done A/B tests and found the default style to be better for metrics.
101.
by dannyw 2mo ago
If only you could use your anthropic sub with a different harness that performs better :( Heck, since Codex is open source, you can just maintain your own personal fork with the things you like (and the things you don't like disabled).
102.
by dannyw 2mo ago
I would probably just give it a few days. Deepseek is usually very good with open weights, they don't necessarily drop immediately, sometimes in a few hours, sometimes in a couple days.
103.
by dannyw 2mo ago
Are you using an API, or running locally? If so, are you running with a quant, or other 'optimisations'? I've been using it via openrouter pretty heavily as my daily driver for the past week and loving it, have never experien
104.
by dannyw 2mo ago
A dashcam could be a safety tool for most people, but it could also be a strategically placed B-cam for a videographer on a shoot. For what it’s worth, dashcam footage can absolutely be copyrightable. The ruling is more about “only humans c
105.
by dannyw 2mo ago
Your highly available system is probably somewhat important, otherwise you won’t have invested in making it HA. While your premise holds for happy cases, when you do have a cascading series of outages, not using exponential backoff is just
106.
by dannyw 2mo ago
It’s been proven that VS Code has product managers who vibe code commits that get merged, including the co-authored with copilot bug. So there’s some problems there.
107.
by dannyw 2mo ago
Modern LLMs are pre and post-trained on order of magnitudes more synthetic data than 'ground truth' or directly human-authored text. But you could say the synthetic data came from a LLM, and the lineage should trace all the way ba
108.
by dannyw 2mo ago
Expert offloading significantly helps with the VRAM capacity. Most MoE architectures have a few experts that are always running; this, the router, KV, and whatever else you have space for can stay in fast VRAM; and the remaining experts can
109.
by dannyw 2mo ago
First, cost optimization directly contributes to how much markup they can charge. Say on average they deliver savings of 20%, that is excess value OR/Stripe can mostly capture. Also, you don't need to mark up tokens much if you&#x
110.
by dannyw 2mo ago
There's no inherent fraud or trademark infringement in advertising/bidding for search terms of your competitors. A classic example is Dominos bidding for 'Pizza Hut', and vice versa. There may be trademark infringement w
111.
by dannyw 2mo ago
I miss the days of camera firmware being flashable. Projects like Magic Lantern bought incredible functionality, plus customizability for your workflow / needs / artistic interests, and those cameras are honestly still _very_ usab
112.
by dannyw 2mo ago
Coding subs are good when they promote usage and adoption of your models in enterprises at API rates. Cerebras is a B2B hardware company. It feels like a distraction: think of the opportunity cost, and resources/headcount not working
113.
by dannyw 2mo ago
Why would they offer a coding subscription and start competing with some of their biggest customers; when they are capacity-bound and companies like OpenAI will take however many wafers that Cerebras sells to them? OpenAI's Sol ultrafa
114.
by dannyw 2mo ago
Cerebras supports prompt caching and has a doc about it. A fairly standard automatic prefix-based implementation with 5min expiry. They do not seem to discount cached input for the self-serve Developer tier. Maybe they do for enterprise rat
115.
by dannyw 2mo ago
One example: Russia Today / RT News. Process: Google made a blog post [1] and cited EU's Council Regulation (EU) 2022/350. The app is still available in Russia on both App Store and Play Store. [1] https://blog.goo
116.
by dannyw 2mo ago
I'm pretty sure Logitech has wired keyboard and mouses that are probably configurable in some way. Or, a webcam for example, or other peripherals.
117.
by dannyw 2mo ago
They’ve announced they’re changing the policy to require ‘verification’ in future Android versions. It will likely be delayed, just like Manifest v3, but eventually it will happen.
118.
by dannyw 2mo ago
The EU likes the power to delist apps by compelling two companies (Apple and Google) to do so. Android is moving to the same, KYC (credit or debit card; no prepaid cards) with play store required for APK distribution
119.
by dannyw 2mo ago
Hard to estimate? Everyone knows the elephant in the room: capable open weight models.
120.
by dannyw 2mo ago
OpenAI says they don't use any API data for training. (there's probably going to be a reply about 'but how can you trust them'; I'm just stating what they say)

More ›