dannyw

17 ms·

451.
by dannyw 5mo ago
The burden of proving two events are related is up to the accuser, and rough time correlation isn't any evidence in itself.
452.
by dannyw 5mo ago
When you have tens of thousands, or hundreds of thousands of employees, your organisational culture and policies inevitably change to limit the impact -- good or bad -- of one individual or a small team.
453.
by dannyw 5mo ago
What makes it a waste of time? A reputable VPN provider that offers a pretty reliable service and has every indication of having a competent security team is worth something in itself; not everyone using Mulled wants to set up / debug
454.
by dannyw 5mo ago
That seems like false equivalence. The overall economy is not zero sum.
455.
by dannyw 5mo ago
You should not be conducting unauthorized penetration tests against third party infrastructure providers without permission. They have processes and systems and usually just wants a heads up of what you plan to test and t the duration &#x2F
456.
by dannyw 6mo ago
Was that Flightfox? If so, I loved using it, helped me save so much money but also time :) It sounds like there’s a problem with having too many flights that are barely full and hence unprofitable. AFAIK the federal gov spends significant m
457.
by dannyw 6mo ago
> The 256 GB/s number is real, but for context, an Apple M5 Ultra hits ~800 GB/s on its unified memory The M5 Ultra has not been even announced. This article appears to be predominately or entirely LLM-produced with little to n
458.
by dannyw 6mo ago
Eh, you're seeing raw thinking tokens. With Claude <x> 4, and I think GPT-5 series, you are no longer seeing real thinking tokens, but "summarized" tokens that are probably highly different to the raw thinking.
459.
by dannyw 6mo ago
In the horror days of IE, I remember having to look up some DirectX filter to properly display PNG images with transparency. It was that bad, and that’s one example of 1000. Some libraries/scripts helped normalise things a little, but
460.
by dannyw 6mo ago
You remember the IE days right? Being a web developer was not fun; and the web was absolutely being held back. Chrome did a lot of things right: per-origin sandboxing, properly implementing web standards, V8, developer tools, and back then
461.
by dannyw 6mo ago
In practice it’s useful too. The local translation in Firefox is quite good, and I love that I can translate pages entirely on my machine; without the contents going to another server. As for Apple foundational models, I think the issue is
462.
by dannyw 6mo ago
This isn’t Mozilla taking a stance against AI. It’s them articulating clear and logical reasons why the proposed API, in its current state, is bad for web interoperability.
463.
by dannyw 6mo ago
I wouldn’t say slightly slower; LLMs are massively useful for software engineering in the right hands. For some personal projects I still stick to the basics and write everything by hand though. It’s kinda nice and grounding; and almost fee
464.
by dannyw 6mo ago
But then they released gpt-image-2, which is clearly SOTA.
465.
by dannyw 6mo ago
Microsoft Azure and OpenAI have been basically hating each other since the beginning, since their incentives were completely misaligned.
466.
by dannyw 6mo ago
For OpenAI, OpenAI direct has always been better; except maybe early 2023-era when OpenAI Platform was not that stable or reliable yet. For Anthropic, it can vary based on model and time. For Opus 4.7, Bedrock is the clear winner in TPS by
467.
by dannyw 6mo ago
So that was our assumption too while building it, but I'm genuinely surprised by how well frontier models can work with large and 'lightly-documented' SDKs. I think a big part of it comes from deliberately exposing lowest-lev
468.
by dannyw 6mo ago
How do you expect the spend & COGS for free LLM inference to be funded? For users who don't want to pay, or maybe can't pay?
469.
by dannyw 6mo ago
If you're interested, for Affinity the way we've built it is through exposing our scripting SDK via MCP. Agents like Claude can write scripts to execute actions, and these scripts can be saved and re-run later, as well have their
470.
by dannyw 6mo ago
I don't think any slop is getting injected into Blender: > Blender Foundation’s mission remains to empower artists with free/open source technology and tools. Yet, we also maintain APIs for individuals and corporations to exten
471.
by dannyw 6mo ago
They don’t have production limits anymore, and can produce and sell as much oil as they have capacity for.
472.
by dannyw 6mo ago
Fortunately there’s plenty of open weight models that are just safetensors and you have a wide variety of providers to choose from, as well as just hosting it yourself.
473.
by dannyw 6mo ago
They can be super complementary! Open weight models can be your everyday standard goto, and frontier models for the harder and bigger tasks. Having some open weight deployment or vendor is also a good thing, because you may have domain spec
474.
by dannyw 6mo ago
A very reasonable and fair markup that is clear and well articulated and not changing (at least so far) on a whim.
475.
by dannyw 6mo ago
There's pretty good consumer protection in the EU. I imagine consumer protection routes probably would apply here. Definitely not really acceptable though nonetheless; you're a paying customer / subscriber that got 'scam
476.
by dannyw 6mo ago
Spend a hour or an afternoon creating your own eval harness with problems or workloads from your private repos or personal projects. Use frontier LLMs to help create the harness and identify problems, but put in the effort to ensure your ve
477.
by dannyw 6mo ago
It’s honestly far better to just ignore SWEBench Verified in 2026. Multiple labs have noted issues with contamination, and achieving high scores require memorisation of what passes the prescriptive verifier; not what is a correct solution.
478.
by dannyw 6mo ago
I agree with your post and the well-studied safety of it in humans at appropriate doses :) but the study you linked is on rats; and so there’s probably another better study to link. I experiment and take a small number of less-commonly-know
479.
by dannyw 6mo ago
This seems designed to provide an easy way for national state actors to get access to anything, since they can just tap and hide a SMS?
480.
by dannyw 6mo ago
When people are moving from other models (e.g. GPT, Gemini, etc) the compute that is previously powering that inference now becomes available. Of course, I'm certainly doubtful that Google would break commits and give OpenAI GPUs to An

More ›