rushingcreek

Many people are talking about price, but I think that the most interesting aspect of this release, by far, is customization.

Any startup can download the weights, tinker with them, and fine-tune them. The real win here isn't necessarily cost, but performance on your data and IP sovereignty. It's a huge win. Kudos to the Kimi team.

This seems like AI washing to me. If the company was already doing well, the AI investments should allow it to grow faster and therefore still be able to give employees bonuses. This announcement implies they view this as a zero-sum investment. Well-run businesses tend not to think this way.

Very cool, congrats! I'm curious how you were able to achieve this given Apple's many undocumented APIs. Does it use private Neural Engine APIs or fully public Metal APIs?

Either way, this is a tremendous achievement and it's extremely relevant in the OpenClaw world where I might not want to have sensitive information leave my computer.

One major point I haven’t seen mentioned is maintenance.

Nvidia GPUs, particularly H100s, have a failure rate orders of magnitudes higher than traditional CPU-only hardware. I myself have accidentally melted an H100 during a large-scale training run.

While it’s trivial to replace a broken GPU in the ground, it seems to be infeasible in space during the life of the satellite. Getting a human or robot “fixer” spacecraft to it would likely cost more than the $30K GPU itself.

Greptile is a great product and I hope you succeed.

However, I disagree that independence is a competitive advantage. If it’s true that having a “firewall” between the coding agent and review agent leads to better code, I don’t see why a company like Cursor can’t create full independence between their coding and review products but still bundle them together for distribution.

Furthermore, there might well be benefits to not being fully independent. Imagine if an external auditor was brought in to review every decision made inside your company. There would likely be many things they simply don’t understand. Many decisions in code might seem irrational to an external standalone entity but make sense in the broader context of the organization’s goals. In this sense, I’m concerned that fully independent code review might miss the forest for the trees relative to a bundled product.

Again, I’m rooting for you guys. But I think this is food for thought.

Congrats on the launch! Definitely a problem I’ve run into myself.

Thank you, will take a look!

Thanks for the specific example -- I'll take a look at this. In the meantime, you can use old.phind.com if you still want the old experience.

I hear you, but why not use something like Google AI Mode or AI Overviews for that? That's pretty hard to beat for simple questions in terms of speed, especially for one-liners.

thanks for reporting!

sorry to hear that -- could you please elaborate?

Oh wow, yeah this one is pretty funny. My attempt gave more reasonable results: https://www.phind.com/search/twinnings-extra-spicy-tea-b1742....

What type of apps would you like to see it make? How does this version of Phind work for it? And thanks for sticking with us :)

A hypothesis here is that well-crafted UI helps you understand/see options for what you don't yet know.

For example, here's an example for a "day trip plan in Bristol" that contains a canonical example (directly based on the query), but also a customization widget that presents some options that you might not have already thought about if you were just doing a text-based followup.

https://www.phind.com/search/make-me-a-day-plan-ac8c583b-ce6...

Thanks for the feedback! The model you use in Phind makes a big impact. Claude 4.5 Opus in Phind gave a better answer than Phind Fast here: https://www.phind.com/search/build-an-interactive-app-showin....

Yep, we heard that feedback loud and clear; the diagrams are a lot less annoying in this new version.

Thanks for the feedback, this is helpful!

Thank you, great to hear :)

hey, sorry to hear that. web search is on by default, but we had some teething issues with it in the last hour. it should be fully fixed now. can you send some links that failed?

Thanks for the feedback, and I agree that it is very much early days for this product category. To be clear, our goal is to make the software specific for an audience: you. What's exciting, though, is that models are rapidly improving at building on-demand software and this will directly benefit Phind. There are still many edge cases, but I think it will get better quickly.

I agree that this answer was a bit wacky. Phind Fast is the fast and free model. Selecting Phind Large, GPT-5.1, and Claude models would be better for a modeling task like this.

The loading issues should be fixed now (as of 11am PST). Apologies for this -- one of our search providers went down right as we launched :(

Thank you! There is a share button in the upper-right corner of the answer page screen :)

Thanks, and great question. The custom Phind models are really key here -- off-the-shelf models (even SOTA models from big labs) are slow and error-prone when it comes to generating full websites on-the-fly.

Our long-term vision is to build a fully personalized internet. For Google this is an innovator's dilemma, as Google currently serves as a portal to the current internet.

Interesting -- could you try with a vanilla browser (no extensions or VPN) please? Preferably Chrome or Safari.

Hi, sorry about that -- we are receiving an HN traffic "hug" spike right now and I'm working on getting that fixed ASAP.

Hi, apologies for this -- it seems to have written a syntax error that it then failed to auto-fix (hence the white screen).

I tried generating your answer again: https://www.phind.com/search/explain-to-me-how-dom-78d20f04-....

Hi, I just clicked the link and it's showing up for me. Could you refresh?

The LLM is phenomenal at figuring out what you want, but it still has to map it to the schema of the tool. So while the job of figuring out the working “rune” is offloaded from you to the LLM, it doesn’t solve the fundamental problem of the available “runes” likely being brittle and insufficient for any given task even when the LLM knows exactly what you want to do.

A fully generative UI with on-the-fly schema would be less brittle because you can guarantee that the schema and the intelligent widget can fully satisfy the user’s request. The bottleneck here is the intelligence of the model computing this, but we are already at the point where this is not much of a problem and it will disappear as the models continue to improve.

I think most software will follow this trend and become generated on-demand over the next decade.