The new Phind goes beyond text to present answers visually with inline images, diagrams, cards, and other widgets to make answers more meaningful:
- "explain photosynthesis" - https://www.youtube.com/watch?v=cTCpnyICukM#t=7
- "how to cook the perfect steak" - https://www.youtube.com/watch?v=cTCpnyICukM#t=55
- "quicksort in rust" - https://www.youtube.com/watch?v=cTCpnyICukM#t=105
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:
- "top 10 Thai restaurants in SF, their prices, and key dishes" - https://www.youtube.com/watch?v=rIQQcDIIHFQ#t=11
It can also perform calculations, visualize their results, and verify them in a Jupyter notebook:
- "simulate 100 coin flips and make graphs" - https://www.youtube.com/watch?v=YP3PZ4MKGCg#t=8
- "train a perceptron neural network using Jupyter" - https://www.youtube.com/watch?v=YP3PZ4MKGCg#t=45
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.
I'm super grateful for all of the feedback we've gotten from this community and can't wait to hear your thoughts!
My prompt,
"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"
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 I've gotten good analysis from other LLM products but they have always been way out of date which makes them useless.
Q: "what are the major economic and earnings events next week and how they can affect SPX volatility and price?"
A: It gave a long and nicely formatted answer with clever visuals, using the right words about FOMC and inflation. Coherent reasoning overall, albeit quite shallow. But the economic and earnings calendars for next week are complete hallucinations, even the dates, so the whole analysis is nonsense:
Wed, Feb 21 Consumer Price Index (CPI)
Wed, Feb 21 EIA Crude Oil Inventories
Thu, Feb 22 Producer Price Index (PPI)
It was enough to give me timelines and risk reward models from a company I had never heard of. This used to take me many hours and a level of obsession that I find incompatible with being social. Verifying is fairly quick, but even if you aren't compelled to do that, now you can at least talk about the company in the associated trading communities and be further corrected or have your understanding challenged.
"Explain why negative numbers are in fact imaginary"
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.
"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."
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.
The response to my second question included a link to an article that suggests all numbers (including natural and whole numbers) are in fact human constructs and may be considered imaginary. That is an enlightening insight that would help us both stop thinking of the words "negative" and "imaginary" as perfectly well defined in our heads. Those words are just tools that can help us convey the most appropriate meaning for the context.
Without the link to the article, that hypothetical conversation probably would not have worked out as well.
Deleted Comment
not really your point, but ??
a decelerating car has negative acceleration, and until it starts reversing relative to its start, it has velocity in whichever direction it started in -- presumably positive if that was your initial frame of reference. of course if you decided positive was the opposite direction from which the car was already going in, well, it started with negative velocity.
also to the GP, if you owe someone a sheep but don't have any, you really do have -1 sheep.
- A decelerating car does not have negative velocity
- "negative velocity" is assuredly nonphysical, which rips the middle out of an argument based on physicality
- Velocity is a vector quantity, as is acceleration. (the steelman'd version of this argument is s/velocity/acceleration)
- The negative isn't physical, it's pidgin for algebra so late middle high schoolers / early high schoolers can imitate physics without learning any of the above
Deleted Comment
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.
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?
Instead, I now just have to prepend to every question I ask:
“Only answer using plaintext, avoid using your code editor and diagram features:”.
(Hilariously this prepend prompt method was suggested by phind itself when I angrily asked “how do I shut off all of these new features?!”)
Which is an additional hassle for me, but so be it.
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.
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.
I’d appreciate a checkbox that I could click to just get a straightforward answer.
(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?)
You can tell it to "only answer using plaintext" there and it will be automatically applied across your searches.
https://support.mozilla.org/en-US/kb/bookmarks-firefox#w_how...https://www.reddit.com/r/firefox/comments/o3yfeo/just_discov...
Did you try including that in your prompt?
Is it a sad state for any tool when one has to specify only wanting the thing they asked for with less verbosity?
Especially when said tool is costly to run, both financially for the service provider and environmentally.
To me, it is, but hey, opinions, ya know?
New Threads Are Public [Enabled By Default]
Do Not Train on My Data [Disabled by Default]
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.
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).
If there are no plans to open the models via API are there plans to support other IDEs (IntelliJ) with the chat feature?
Please let us know!
I think building comparison tables is one of my favourite things to do here. Saves me considerable amounts of time and saves me from my biases to some extent.
I think the new Mermaid support is a great idea. It sure is handy that, before LLMs were even a thing, we were already collectively working on so many textual, readable languages to describe things like this! I am going to try to use it to create some architectural diagrams by adding requirements one by one.
However I did find myself wondering how crucial really were the model changes?
Imagine trying to implement these features as a wrapper around frontier apis and no 70B bespoke model.
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.
How much less would this simple approach leave us with?
We write more about this in the technical model blog post: https://www.phind.com/blog/phind-2-model-creation.