Topfi

30 ms·

421.
Ink and Switch – An independent research lab exploring tools for thought(inkandswitch.com)
5 points by Topfi 10mo ago|0 comments
422.
I Know Why Lying about AI Water Use Is So Easy [video](youtube.com)
4 points by Topfi 10mo ago|0 comments
423.
by Topfi 10mo ago
I very much like the direct demo, similar to how Groq showcased their offerings initially. Would love if they shared more about their custom hardware and model training, as it stands this is far less information than Cerebras and Groq offer
424.
by Topfi 10mo ago
The name of the event where this took place was "Autonomy Visualized". Can we make securities fraud properly illegal again or how many more years of this will we have to endure?
425.
Anthropic's philosopher answers your questions [video](youtube.com)
3 points by Topfi 11mo ago|0 comments
426.
OpenRouter Broadcast(openrouter.ai)
35 points by Topfi 11mo ago|12 comments
427.
by Topfi 11mo ago
Isn’t it post installation still updated via Windows Update as they said (force end-users to update the framework)?
428.
by Topfi 11mo ago
> [...] quickly pushed it to the market first to get patents and get all the profits [...] Beyond what others have commented already, especially on obesity and cardiovascular disease, I have to correct this specifically, because it is a
429.
by Topfi 11mo ago
Sorry, but can you point me to what part of the system prompt here would/could be responsible for causing MechaHitler? I have yet to see anything in the prompt they claim to have been using that would lead to such output from models by
430.
by Topfi 11mo ago
There is no paid tier on Antigravity currently.
431.
by Topfi 11mo ago
> If you ask Claude to fix a test that accidentally says assert(1 + 1 === 3), it'll say "this is clearly a typo" and just rewrite the test. Codex will rewrite the entire V8 engine to break arithmetic. Honestly thanks, in t
432.
by Topfi 11mo ago
Tested it for roughly two hours now, far from ready for prime time, very buggy and clearly just quickly build on Windsurfs already rather issue laden code base. Essentially a less well thought out imitation of Trae's Solo mode, added i
433.
by Topfi 11mo ago
Had the same issue, have been able to sign in finally using a Google Cloud Identity (former Workspace) account by changing my IP via a VPN to Singapore. No idea why, but that worked. Tried a few other countries too, but only had success wit
434.
Gemini 3 Pro Preview on OpenRouter(openrouter.ai)
1 points by Topfi 11mo ago|0 comments
435.
by Topfi 11mo ago
Google has been doing that for a while.
436.
by Topfi 11mo ago
Max context is unchanged from 2.5 Pro at 1M token.
437.
by Topfi 11mo ago
Additional context from AI Studio including pricing: Our most intelligent model with SOTA reasoning and multimodal understanding, and powerful agentic and vibe coding capabilities <=200K tokens • Input: $2,00 / Output: $12,00 > 2
438.
by Topfi 11mo ago
Additional context from AI Studio including pricing: Our most intelligent model with SOTA reasoning and multimodal understanding, and powerful agentic and vibe coding capabilities <=200K tokens • Input: $2,00 / Output: $12,00 > 2
439.
Gemini-3-pro-preview available on AI Studio(aistudio.google.)
2 points by Topfi 11mo ago|4 comments
440.
Gemini 3 Pro Model Card(pixeldrain.com)
431 points by Topfi 11mo ago|2 comments
441.
by Topfi 11mo ago
> Apple wouldn't just sacrifice the entry-level MacBook product category and I'm not even sure about that - the look-and-feel of a "display with attached keyboard" (i.e. Thinkpax X1 Tablet-style) is vastly different f
442.
by Topfi 11mo ago
Historically, cannibalizing has always been the right choice when it comes to such things. That was a major point of the first iPhone, that it was a full replacement for your iPod, which was instrumental in its success. All this thinking do
443.
by Topfi 11mo ago
This is a great way to demo and I wish other labs took a page out of Anthropics book.
444.
by Topfi 11mo ago
> Yahoo has shown that even with immense brand recognition people will still switch if the alternative is really better. But it didn’t, Google users beyond a few initial super users who switched early on in the late nineties/early t
445.
by Topfi 11mo ago
Gaming on PC or using Adobe software is very far removed from the average in 2025. Same if you use any Office application after work hours. Heck, even Canva is, the vast majority of users are not creating on their systems, solely consuming
446.
by Topfi 11mo ago
You don't have to take my word for it. Open the three next to each other, type in a search query. Only one will provide part of the results as a wildly inaccurate "overview" that somehow manages to link to sources telling you
447.
by Topfi 11mo ago
Kagi gives you free searches and you pay for Google with you data/advertisements so it costs something too, but you knew that already. And both provide more reliable results, so what are you trying to say?
448.
by Topfi 11mo ago
Both DuckDuckGo and Kagi are, purely focused on result quality in 2025 far better than Google currently, but that does not matter if one has no need beyond the most basic results. And much of the MS Office suite is actually a worse copy of
449.
by Topfi 11mo ago
Keep in mind how severely the quality of Google search results or e.g. any consumer facing piece of Microsoft software (Windows, Office, OneDrive, etc.) have deteriorated to the point it has far transcended the more nerdy corners of the web
450.
by Topfi 11mo ago
In fairness, that is for their model, not their brand. In this industry, the mind share and buy in OpenAI has is still second to none, even when Gemini 2.5 Pro served the vast majority of users far better than the OpenAI models, the latter

More ›