Hi HN! Michael here. We've spent the last 6 months rebuilding Phind. We asked ourselves what types of answers we would ideally like and crafted a new UI and model series to help get us there. Our new 70B is completely different from the one we launched a year ago.<p>The new Phind goes beyond text to present answers visually with inline images, diagrams, cards, and other widgets to make answers more meaningful:<p>- "<i>explain photosynthesis</i>" - <a href="https://www.youtube.com/watch?v=cTCpnyICukM#t=7" rel="nofollow">https://www.youtube.com/watch?v=cTCpnyICukM#t=7</a><p>- "<i>how to cook the perfect steak</i>" - <a href="https://www.youtube.com/watch?v=cTCpnyICukM#t=55" rel="nofollow">https://www.youtube.com/watch?v=cTCpnyICukM#t=55</a><p>- "<i>quicksort in rust</i>" - <a href="https://www.youtube.com/watch?v=cTCpnyICukM#t=105" rel="nofollow">https://www.youtube.com/watch?v=cTCpnyICukM#t=105</a><p>Phind is also now able to seek out information on its own. If it needs more, it will do multiple rounds of additional searches to get you a more comprehensive answer:<p>- "<i>top 10 Thai restaurants in SF, their prices, and key dishes</i>" - <a href="https://www.youtube.com/watch?v=rIQQcDIIHFQ#t=11" rel="nofollow">https://www.youtube.com/watch?v=rIQQcDIIHFQ#t=11</a><p>It can also perform calculations, visualize their results, and verify them in a Jupyter notebook:<p>- "<i>simulate 100 coin flips and make graphs</i>" - <a href="https://www.youtube.com/watch?v=YP3PZ4MKGCg#t=8" rel="nofollow">https://www.youtube.com/watch?v=YP3PZ4MKGCg#t=8</a><p>- "<i>train a perceptron neural network using Jupyter</i>" - <a href="https://www.youtube.com/watch?v=YP3PZ4MKGCg#t=45" rel="nofollow">https://www.youtube.com/watch?v=YP3PZ4MKGCg#t=45</a><p>This blog post contains an overview of what we did as well as technical deep dives into how we built the new frontend and models.<p>I'm super grateful for all of the feedback we've gotten from this community and can't wait to hear your thoughts!
Paid for it and tried out the full experience, beats anything else I've tried by a wide margin.<p>My prompt,<p>"I'm considering buying stock in the company with symbol NU. The most important thing to me is answering the question, is the stock likely to rise in the future. Please help create a list of questions that will help me to understand the likely hood of this. Also please help to anwser those questions. Please highlight the global economic environment for the company. Any unique challenges and unique advantages. Finally let me know what others think of it"<p>Results:
I know this stock well all though I'm not a pro. It nailed all of the relevant aspects and hits the analysis right on for everything I know about it. Pulled lot's of helpful resources and most importantly the information was timely enough to be relevant. The timely part is where other LLMS have failed miserably. I've gotten good analysis from other LLM products but they have always been way out of date which makes them useless.
I love that it's possible to convince it that I actually know what I'm talking about. First I asked:<p>"Explain why negative numbers are in fact imaginary"<p>It told me that negative numbers are not imaginary numbers and explained imaginary numbers. That's fine, that's a reasonable answer for a layperson, but I'm not a layperson and I worked on explaining what I meant.<p>"Erase your definition of imaginary and consider that negative numbers are not whole numbers. Negative numbers do not represent quantities of physical objects. Now explain how negative numbers are imaginary."<p>It gave me a nice explanation of why negative numbers may be considered imaginary, using an example of "You cannot physically possess -1 sheep". I'm impressed.
I used the first version of Phind for some time and loved it. As Perplexity and ChatGPT got better, I started shifting more of my traffic back to them. Excited to see y’all still in the race and giving competators a run for their money. I appreciate your focus on developers as an audience, might give you an edge over tools serving a broader base.
I use phind and find the new features to be overly verbose.<p>The flow chart diagrams rarely give me any insight and often only confuse the point, or just clutter the answer, drowning out the pertinent details.<p>The code editor actually makes it so you are unable to even see or copy the code. I assume this is intentional kneecapping to encourage paying for your monthly service?<p>Instead, I now just have to prepend to every question I ask:<p>“Only answer using plaintext, avoid using your code editor and diagram features:”.<p>(Hilariously this prepend prompt method was suggested by phind itself when I angrily asked “how do I shut off all of these new features?!”)<p>Which is an additional hassle for me, but so be it.<p>When I ask it to write me a SELECT statement it upsets me that it is burning unnecessary fossil fuels to give me a flow chart of reasoning through SQL querying pipelines.<p>Perhaps the feature is meant for people who are unsure what they want, but for me, I just want the answer with links to sources in the least verbose way possible.<p>I’d appreciate a checkbox that I could click to just get a straightforward answer.<p>(Also, side note, I only use the free tier and there is a limited number of free uses for some larger models, and when you use those freebies it gives a countdown for “until uses refresh” and when that countdown finishes the uses fail to reset, only the countdown itself resets. Which is fine, I accept that I only use the freely offered model, previously “instant” currently “70B”, with its clear flaws, but it’s just another frustrating UI feature that seems to fail to live up to its promises so I am, again, just confused why it’s there?)
Settings > Profile > Preferences<p>New Threads Are Public [Enabled By Default]<p>Do Not Train on My Data [Disabled by Default]<p>The first one seems problematic -- does it mean what it seems to suggest? all conversations are "public" ?? Do users know this going in? i am afraid they could be pasting lot of private or confidential stuff without realizing this.
Hi Michael, thank you for all the hard work that goes into the Phind models and congratulations on the new UI. Been a paying customer since first finding Phind here on HN in September.<p>Question: are there any plans to allow access via API to integrate the Phind models with the Continue plug-in (would also love to integrate into my personal RAGs)? Mostly using IntelliJ and integration would be awesome. Do have the VS Code plugin setup and use that when needed. Also running the smaller Phind models locally to use with Continue, but that only gets me so far without needing to open the real UI. If the API opened both the 405B for chat and the 70B for auto complete would be a big step in gaining more paying customers (IMO). No need to open the other models as those can be done with other API keys (if one wanted).<p>If there are no plans to open the models via API are there plans to support other IDEs (IntelliJ) with the chat feature?<p>Please let us know!
On a positive note, this is a nice look at the future and a direction that existing experiences seem likely to evolve towards.<p>However I did find myself wondering how crucial really were the model changes?<p>Imagine trying to implement these features as a wrapper around frontier apis and no 70B bespoke model.<p>Starting with a user query, we could ask for relevant svg diagrams, fit instruction steps or guides into layout templates, or even filter photos as the demo shows.<p>How much less would this simple approach leave us with?
Oh hey, I used phind a lot a while back and really enjoyed it. Sort of fizzled out of it for other stuff a few months back, but this looks pretty exciting! The visual explanations are great. I've been getting gemini to output ASCII diagrams to explain things, will definiely be checking this out.
I asked "What is Atproto that bluesky uses" and it does a Technical Architecture illustration and it is wonderful. I think that is probably your new value and might be a good idea to highlight it somewhere at the top rather than showing the whitepaper.
I still don't understand why these models can't be more "trustworthy", but I also don't understand the theory, I'd love to hear what you all think about this.<p>I asked it a question that's pretty subjective and culturally specific, and I appreciate that I got a reasonable answer back. The question was "should I?" and the answer was "definitely, don't miss it" in three different ways. However, I found that the literal sources it quoted didn't have the same opinion it expressed to me so convincingly. I asked a clarifying question and it goes "okay so I read the material, and it actually says it's optional".<p>So why not read the material? I wonder if it could even embed the website in the results, giving the website the traffic and ad space. I wonder if a meta browser is a better product for these tools.<p><a href="https://screenshot.click/13-56-232ze-p3nzf.png" rel="nofollow">https://screenshot.click/13-56-232ze-p3nzf.png</a>
Congratulations on the impressive Phind 2 rebuild! The visual answers and multi-step reasoning capabilities look fantastic. I noticed that currently only the Phind series models support image uploads, while Claude 3.7 Sonnet in Phind 2 doesn't have this functionality yet. Do you have plans to enable image upload support for Claude 3.7 Sonnet in the near future? Being able to use Claude's multimodal capabilities within Phind would be a great addition, especially considering how powerful Claude's latest updates have been. Thanks!
Hi Michael, the range of unique interactions and little pieces of visual feedback for users in your demos is super impressive, and as a designer, I hope you don't mind if I shoot my shot.<p>I'm ending a MIT xPro course on designing AI products in about a week, and I've been looking for places to put all my learnings to work, might your team have some room for a coding/prototyping UX developer/designer who has spent a lot of time thinking about AI-oriented HCI?<p>I think in 8 weeks of a co-op / or even part time basis, I could build out a ton of ideas you haven't gotten around to defining yet, but want to get a feel for the UX of. I'd love to hear from you (my email in in my profile).
Well done. After running a few sample queries I seem to get about the same quality answers as I'd get from perplexity, although with more images included. Overall great progress and a solid competitor to other AI search engines I'd say.
I liked phind and subscribed for quite a few months. I used it for both search and for coding. imo, chatgpt search is terrible...It constantly relies on spammy sources.. phind resultsback then tended to find the quality sources, letting it give you quality answers.<p>But I started wondering if it was phind was a nearly dead company: Almost no presence in terms of news, tweets, etc. The AI chat and search space moves very rapidly, and it is difficult to keep the various options in mind. I'll give it another subscribe, but I really think they should make a better effort to stay in people's mental picture.
This is amazing. I paid for Phind many months ago for coding QnA. However, this is a much better product and even beats Perplexity, ChatGPT premium, etc. Congratulations to the Phind team!
As far as a search engine goes, Phind beats Goodle 10x times in programming topics.<p>Google Search is extremely dumb nowadays as it promotes ad-banner-ized results and Reddit-style drama posts. Those things have a close to zero value, but Google shows them in top results to earn money and imaginary "engagement". The trend is getting worse every year. I predict that one day AI assistants like Phind will smoke conventional search engines out of the water in every imaginable area.
I remember when I was in school and our religion teacher told us that 200 years ago there were still people that knew almost everything that was known to humanity back then. This wasn’t possible anymore.<p>However now I feel like it’s possible again.<p><a href="https://www.phind.com/search/cm73yk4wc0002336u6brx4jp1">https://www.phind.com/search/cm73yk4wc0002336u6brx4jp1</a><p>It helps greatly to learn new topics.
Phind does one thing and it does it really well. I use it because it has a real, positive impact on my growth. I love the fact that it is not Jarvis, but just a really helpful library assistant. Please don't lose the focus! Stay awesome.
It's better than perplexity on certain tasks! Pretty impressive!<p>Feedback: the mermaid charts are a bit annoying when they are not needed<p>Our of curiosity, what web index do you guys use right now and does it make sense to build your own index at some point?
Gotta respect the hustle of the guys. Some work colleague recommended it to me over a year ago but I stopped using it after a couple searches, then never heard from them again after perplexity exploded. Hope it works out for them!
Wow, this could conquer the "normie" crowd that likes using ChatGPT to answer search-like prompts. The UI is so much cleaner than any other alternative and the outputs are like I'm reading an article about the subject.
In your "how to cook the perfect steak" video [1] there's a picture of various doneness levels of a steak. It's a fantastic picture. The creator of that picture will get jackshit from this. Phind gets value, the user gets value but the creator does not.<p>You're hyperlinking to the source, which is nice. But there's no reason for the user to click through so it won't really help the creator. The upshot of all this is that the open web will have less incentives for content creators. Someone's got to create new data for the AI borg. In future, these creators are less likely to be independent bloggers/photographers. Perhaps biased media outlets will be the only ones with the incentives to create.<p>[1] <a href="https://www.youtube.com/watch?v=cTCpnyICukM#t=55" rel="nofollow">https://www.youtube.com/watch?v=cTCpnyICukM#t=55</a>
I've been trying to use tools like these for data aggregation, analysis, and finding trends. One major issue I experience with almost every product is I will tell it to "sort by column x" and it either doesn't do it, or will state that it did it when in fact it did not do the sorting. Check out this query: <a href="https://www.phind.com/search/cm7444is700011y6rtzk6l5th">https://www.phind.com/search/cm7444is700011y6rtzk6l5th</a>
Can you add the ability to ask questions via GET requests? This is needed to add a site search keyword into the browser. Make it work only when logged in, if you are afraid of bot requests.<p>There is also a standard to add it automatically ( <a href="https://stackoverflow.com/questions/38670851/whats-a-script-type-application-ldjsonjsonobj-script-in-a-head-sec" rel="nofollow">https://stackoverflow.com/questions/38670851/whats-a-script-...</a> )
when someone gives a link: <a href="https://www.phind.com/search/cm73yk4wc0002336u6brx4jp1">https://www.phind.com/search/cm73yk4wc0002336u6brx4jp1</a><p>anyone else can edit those links and have it persist.<p>Like, when i clicked that link it was 3 prompts trying to get phind to prove 1=2. Now it's about the TV show weeds and down quarks.<p>I don't want to expend any more effort to report than this because linkedin wanted me to pay money to send a message to whoever was claiming to be from phind in comments<p>so i send that link to a friend. it shows them 1=2 convo, they edit it. i hit refresh, it shows their edits.<p>This isn't just some "oh it's the same IP" or "cookies" or whatever. My cellphone isn't on my wifi, and my friend lives 1100 miles from me on a different provider.<p>Furthermore, i am unsure of what the OP originally had in that link, because they said it showed that AI could help teach new concepts, and this looks like a joke meme prompting. So who knows what's going on.<p>A guess? the links are SESSION links instead of CONTENT LINKS. A session link allows edits and a content link is just a hotlink to render the content that the session created.
The pricing on these products is frustrating. To even try out the Phind-405B model I have to pay $20 and then I get unlimited queries which I never asked for? For a product I don't even know if I'll ever use again, give me usage based pricing.
Good to see non-mainstream companies like Phind find interesting and creative ways to be competitive. If I was the founder of a company that wasn't an Open AI wrapper (i mean this as matter-of-fact, not in a negative way), I would look to Phind's launch of this product as inspiration, especially considering how massive the warchests are of my mainstream competitiors.
Since the homepage is a prompt to interact with the bot, I directed my questions about possible plans to it.<p>It said: "Up to 10 uses of best model per day"<p>When asked what one use means, it kept checking the docs and told me: "Each interaction with Phind's models counts as a usage".<p>When asked to clarify, it ultimately said: "I cannot confirm exactly whether this would count as your third usage today, as I don't have direct access to your usage statistics. To check your current daily usage count, you can visit your settings page at <a href="https://phind.com/settings">https://phind.com/settings</a>. This will show you your exact usage count for the day and remaining available uses based on your subscription tier."<p>Since I'm not a user and am inquiring about details before subscribing, I had hoped the bot had some more information on that.
The new version feels great.<p>What I always liked the most about it though, is how fast the webpage loads up when I reach out for it. The entry barrier feels lower than the competition, so I usually ask Phind first, then eventually try with other engines.<p>Keep it up guys, I'm a big fan.
These graphs were available to me for some time now but I have not really found them all that interesting. For some reason, I had to wait for the diagram to render and then for the text to slowly render.<p>To my poor understanding of LLMs, when the diagram was slowly created - the text behind it should already "be there" and should have been displayed immediately after the diagram but this was not the case. Also, often the slowly drawn diagram was only presenting my rather clear (for LLM) prompt: "i did this, when situation was this and that, and than this happened; question: why the result was A and not B?"<p>I found myself falling back to Claude more often than not over using Phind 70B and 405B models. I found it kind of... more gimmicky than useful.
This is awesome, I was thinking of creating an agent to generate spec and sequence diagrams (using Mermaid), but this takes it to a new level being generic. I tried a question about how does LLM post training work and it made a nice workflow about it.
Asked a tricky tax question concerning AMT on LTCG. Phind 70b answered quickly and correctly. ChatGP-4o thought longer and answered incorrectly. o3-mini gave correct answer, but spent a lot of time thinking. Impressive.
Probably a loaded question but how has your 70B model stood against the test of time? I would expect models like Qwen2-72B to outperform yours noticeably? (I get why you can’t simply fine tune that due to the restrictions on that license)
I'm on Phind's free plan, and have an OpenAI API account, which gives me access to the suite of OpenAI's models on their playground. I go to Phind's free option first and get what I need then, 80% of the time.
I have been using Phind for building SQL queries and it's actually great to learn more about SQL, and the results are mostly correct. Just try to ask it to build some advanced query then you discover you don't truly know SQL.
Was curious about how this would perform for something I was actually interested in this morning. I live in Canada and want to sell my PS5 to GameStop, and I was wondering what accessories to include. Overall it performed quite poorly answering this. First, it claimed GameStop doesn’t operate in Canada (wrong, EB Games rebranded awhile back), second, it thought I had to “charge” my PS5 first. It then listed some components to bring in but was missing the HDMI cable requirement. Overall not great for this particular query.
So I decided to run a simple comparison with OpenAI's mainstrem model based on a vague question which would require follow-ups.<p>----<p># Phind 50b<p>User: how do I a whole coca-cola bottle<p>Assistant: I notice your question isn't quite clear. Could you please specify what you'd like to know about Coca-Cola bottles? Are you interested in:<p>How to identify/distinguish different types of Coca-Cola bottles?
The history and design of Coca-Cola bottles?
Something specific about collecting Coca-Cola bottles?
Or perhaps something else entirely?
Once you clarify your interest, I'll be happy to provide detailed information tailored to your specific question!<p>User: I just accidentally a whole coca-cola bottle and how I need to it<p>I notice your message isn't quite clear. Could you please rephrase what happened with the Coca-Cola bottle? Are you saying you:<p>Accidentally dropped/broke a bottle?
Spilled the contents?
Found an old bottle somewhere?
Something else?
Once you clarify what happened, I'll be able to provide specific guidance on how to handle the situation safely and effectively.<p>----<p># gpt-4o with a custom eigenprompt<p>User: how do I a whole coca-cola bottle<p>Assistant: do you mean:
1. consume an entire coca-cola bottle in one go? (drink it fast, chug it, etc.)
2. acquire a whole bottle? (buy, steal, manifest from the void)
3. destroy a whole bottle? (smash, melt, vaporize)
4. become a whole coca-cola bottle? (metaphysical transformation, brand loyalty taken too far)<p>clarify and i will give you the optimal strat.<p>User: I just accidentally a whole coca-cola bottle and how I need to it<p>Assistant: okay so if you accidentally the whole coca-cola bottle, and now you need to it, then we need to determine:
1. did you drink it too fast and now your stomach is rebelling? (solution: burp aggressively, lie down, accept fate)
2. did you drop it and now there’s sticky chaos? (solution: paper towels, baking soda, maybe a flamethrower)
3. did you somehow ingest the actual bottle? (solution: seek medical assistance bc that’s wild)
4. did you turn it into an existential problem? (solution: embrace the absurdity, become the bottle)<p>gimme specifics, and i’ll troubleshoot this disaster with you.<p>----<p>Overall, I think that phind model is clearly inferior — it shows less imagination in the face of uncertainty and certainly much less creativity and humour.
I asked one question where I expected a nice non-trivial visual diagram and my expectations were met and exceeded:<p>> How to make an enchanting table in Minecraft from scratch?
As an occasional user, what I really liked in your previous version was that Phind always gave the list of sources.<p>So my workflow was :<p>- ask a question<p>- quickly read the given answer<p>- go to the main sources to decide if they are trustworthy or not<p>I noticed that the new version does not always give the sources, even when I ask. Is this intentional or do you plan to revert this in the near future ?<p>It also solves the problem of not reducing the traffic to the content creators you use.<p>Anyway, congratulations for making this tool and offering a free plan to test it.
The UI is brilliant !<p>1. I like the way sources are listed on the right panel, that makes them easily visible as i scroll through the answer.
2. The rendering of the answers using rich visual components is nice to look at. I personally hate information dense reports that ChatGPT spews, which isn't easy to consume quickly.<p>What I miss though is iterations on the answer, as in, how do I specifically zoom into one aspect of the answer, and then easily zoom out and continue.
This has also happened with Perplexity and R1 to me, but it was just one word. Suddenly the text goes into another language. I just clicked homepage’s suggestion pill “Ukraine war latest”<p><a href="https://paste.pics/STGU6" rel="nofollow">https://paste.pics/STGU6</a>
Thank you for innovating. I am a subscriber and use phind daily and I've noticed the improvements.<p>Is there any way I can provide feedback or contribute feature ideas/requests? I've got some things I'd like to share if your team is prepared for the feedback.
I asked for a map of Paris with all the arrondissements (since there's no way to get that from google maps) and it showed me a diagram instead of a map. Chatgpt at least gives me a link to the wikipedia page with the map image.
I haven't tried Phind, but the steak example has a lot of errors. A good half of them are wrong, but for example, it says not to flip too often. Actually, flipping frequently improves even heat distribution in the steak.
Wish we could have PowerPoint slides generated by phind. It already can generate the SVGs/PNGs. It's such a no-brainer to extend that to slides!<p>Maybe one day we'll have short clips generated by AI, something like 3Blue1Brown videos...
I was a bit fan of phind.com for a while and I loved it. But I've been using duckduckgo's AI offering since its UI is a little smoother/faster/easier to use. Has anyone found any good reason to use phind over ddg?
I wish there was a way to use this via an API instead paying for yet another subscription…<p>I’m going to pass on the niceties shown here until I can try them out paying for what I use instead of an API..
I am very impressed and I am getting great results for queries like "Show me how to use langchain in python".<p>However I am disappointed that when I provide a url it can not read the page. Given that this is a search engine I would expect it to be able to read any public URL I provide it. For example I attached a PDF of my resume and provided a link to a public job description and asked it to generate a cover letter tailored to my experience for this position. This is something I have done with easy success with ChatGPT GPT-4o, but Phind throws its hands up. :(
Can someone who has used either You.com or perplexity recently tell me how they compare? The phind landing page is uninformative. I would encourage them to use it to explain their value proposition relative to the alternatives, which are well-funded. I could not find pricing details; is there no paid plan?<p>From their videos above I gathered that they format their results and incorporate media better?