By having to defend your (thesis)/work like this, the whole piece is getting lifted into academic heights in a way, so you could as well keep calling its result and process research :)
What description would itself come up with, BTW?
When you anwer with "I agree, LLMs have biases.", I immediately suspect that to be an LLM calming me after correcting it, though. So, the world has definitely changed and we might need to allow for correcting the broadness of words and meanings.
After all you did not write thesis, scientific research or similar and I remember it being called researching when people went looking up sources (which took them longer than an agent or LLM these days). Compressing that into a report might make it a review, but anyway. Great that you assembled a useful work tool here for some who need exactly that.
I'm very curious. As it's main purpose is built to really search and decide balanced/unbias sources and not really have an opinion, what the result would be of such a question. I'm curious if it will give an answer on this. I just gave it the request:
"I am struggling what to call this other than "Deep Research tool" as really it is looking online and scanning/synthesizing sources (that's you, by the way!). With that in mind, someone suggested "literature review" but it makes me think of books. I wonder if you can see what this kind of "research" is and suggest a name to describ it based on all the information you uncover on what good research looks like."
Let's see how it gets on...
Also, something I think about a lot (you sound like a deep thinker!) - when we discover something that is untrue, can it make it true? (purely hypothetical thought)... if 1000 people were told coffee was bad for them, does the mind-body connection take over and amplify this into reality. We are certainly in interesting times!
Well, it seems it uses the word "research" quite a lot for what it's doing (first paragraph and last one with "deep research") but does choose some artificially constructed one as the result.
Personally I had similar results when searching for known terms for certain concepts that I didn't know the name for. And usually I had to guide the process to find the actual expression used in the domain (it usually made up a lot of fancy and well-fitted names itself). And sometimes it helped to gonback and change the query. Harder if you have to wait that long, though :)
So.. I guess I would trust its process (mentioning "research" a lot) more than its chosen result :) Not sure if it you wanted it to be a synthesizer or rather an assistant. I'd go with what is closer to your intention or what you think it resulted in in the end.
An interesting observation might be that guiding that (I don't dare to say it, but anyway) research process might still be an important part and the network's self-evaluation might not be as good as one would need it to be at this point in time. I'd guess it's about adding ones personal judgement early to not end up at the wrong spot after a long processing time. Then again you are using a certain selection of simple and more complex models, so I can't say if there might be a way to have that kind of harsher judgement that one would apply oneself emulated in the proces and what side effects might come form that choice of models you made.
In the end I was just surprised by the number of picky replies your post got, so I just thought that you discussing the perfect description with the LLM might make be a fun solution. Personally I am a big fan of interactively talking to LLMs at the moment though, so I might be the wrong guy to know your intended audience and use case enough. Just couldn't see how using the term "research" would be the problem.
Thanks for the feedback, hoc. I am already re-coding the whole thing to use more of an Agents/crew style approach where certain roles can be assigned to critique the research, fact-check, etc.
I re-built with CrewAI but was limited so now doing the same again but using AutoGen.
The criticism is important to this, as is any research project. That's why good research works, if it is challenged. I'm implementing that "challenging" directly into the LLM as separate agents.
Balancing token efficiency with regular check-ins and critique is important, and something I want to ensure the user has control over.
Nearly to version 0.1.3, and much to learn still.
I loved that use case! I will keep testing, refining and hoping others will do until we get there. Persistence is what matters, to try our best and strive towards collaboration, openness and accuracy.
Obscurity is dispelled by augmenting the light of discernment, not by attacking the darkness.” - Quote by Socrates.
I updated the Readme now, to describe as "CleverBee: AI-Powered Online Data Information Synthesis Assistant" and put emphasis on the synthesis.
Also, put a new section in place:
What cleverb.ee is not
Cleverb.ee is not a replacement for deep domain expertise. Despite explicit instructions and low temperature settings, AI can still hallucinate. Always check the sources.
That also is similar in a sense to a typical human bahavior of "rounding" a "logical" argument, and then building the next ones on top of that, rounding at each or at least many steps in succession and bacically ending up at arbitrary (or intended) conclusions.
This is hard to correct with a global training, as you would need to correct each step, even the most basic ones, instead. As it's hard to convince someone that their result is not correct, when you actually would have to show the errors in the steps that led there.
For LLMs it feels even more tricky when thinking about complex paths being encoded somehow dynamically in simple steps than if there was some clearer/deeper path that could be activated and corrected. Correcting one complex "truth" seems much more straightforward (sic) than effectively targeting those basic assumptions enough so that they won't build up to something strange again.
I wonder what effective ways exist to correct these reasoning models. Like activating the full context and then retraining the faulty steps, or even "overcorrecting" the most basic ones?
The promptly emerging Excel-Schmerz reminds me that the author avoided the use of floating point numbers in the provided example for a reason.
In german Excel it's might be ok due to its original focus on numbers, but replacing decimal "." with the german "," also forces you to use a semicolon everywhere where you would use a comma (function, vector/array notations). To me the most annoying issue with localized programming languages. And consecutive ";;;" just looks awful. And what would happen to command/line endings...
BTW, Apple once tried to translate content-specific parts of their AppleScript language (like Dialogs etc) but in the process also hit some enumerations which often sit at the border between programming language and content. Big desaster in certain edge cases.
Anyway. Of course I like that the effort was made. There are famous Asterix comic books translated into all kinds of languages and dialects. I'd really like to see great localized coding languages, but I guess they will have to avoid line ending semicolons and only use integers :)
> In german Excel it's might be ok due to its original focus on numbers, but replacing decimal "." with the german "," also forces you to use a semicolon everywhere where you would use a comma
The best part is that sometimes you get new numbers.
Excel is showing to a user what AI can do to a document: destroy it.
It somehow feels that one has to come up with a tailored strategy for each CVS import that contains decimal numbers, dates and/or other data with dots in them. Most of the time I end up regexing my way out of the failing (and different) attempts the different build-in options in Excel deliver. LibreOffice is better, Apple's Numbers partially worse in that regard. Funny to see these things in that state in 2025.
Discussing number data with current LLMs also does not seem right at the moment. Not sure if simple column-specific conversion tasks might work.
Is this really the action/messaging pattern that Next, Be and then Apple and probably others used about 30 years ago, or did I miss something here.
It was useful in a way but basically evolved into interface-based controller patterns due to the needed complexities and the whish to keep the basic underlying design pattern. So I'd expect to see lots of enhanvement requests once that box was opened :)
There was an early Java UI toolkit by some browser vendor (IFC by Netscape, probably) that allowed to string action elements together like that.
You thought web was 15 years behind, but it quickly catches up to 30 years old tech!
News like this make me feel sick. I thought it could become something usable in my lifetime, but I’m knee osteoarthritis years old and it still reinvents poor man’s target-action. Sad clownage.
For headphone audio it's plenty to just PWM, inductance of the headphones and mechanical low pass characteristic of the speaker itself will mostly do the rest.
If you want to smooth the things out because you target a line input, simple LC filter will work just fine here.
Indeed. fast PWM is a great approach and fits in well with the analog conversion side effects.
I still find the principal question of "if I need additional components anyway, should I just get the more complex ones" interesting. It's located somewhere between quality, price, the definition of simple and optimal, as well as one of ones personal (or this week's) perspective on the art of the craft.
The projects is not a discrete player that basically pushes SD card data via a few digital standard ICs out as PWM, it uses a pretty sophisticated MCU after all. So, is a I2S DAC the logical solution or the fast PWM, I still find it hard to decide, especially when i have to order (even the simple or
mechanical) parts anyway, or should I have a few I2S DACs in the drawer with the Pico these days.
Not sure if I can bring that point across. It's more about what are your default parts these days. I always have a few i2c OLED displays in that drawer, and the project uses one, too, thirty years ago it might have been resistors, LEDs and a few logic ICs instead. So I wonder, would I go discrete (or "simple") for that particular part of the solution for a particular reason or not. And also, why do I still have no I2S DACs stocked in that drawer, while I have display, sensors etc.
So maybe it's just about audio being a somehow ignored interface in (my) embedded tinkering. Or it's more general and about a modernized selection of stocked parts and using those more naturally while still being aware of the simpler solutions. I use pretty complex (but inexpensive) I2C sensors these days, after all.
Ok, many words on that small part of the overall decision making...
We did something like that with an Amiga bouncing ball demo (on linux or Macs) back in the day (two decades ago, probably)
I think it was using a wide master screen that the terminals connected to via vnc, displaying different parts of the master screen. Only three or four screens wide, though, and not optimized for user interaction. But a live app nonetheless.
What description would itself come up with, BTW?
When you anwer with "I agree, LLMs have biases.", I immediately suspect that to be an LLM calming me after correcting it, though. So, the world has definitely changed and we might need to allow for correcting the broadness of words and meanings.
After all you did not write thesis, scientific research or similar and I remember it being called researching when people went looking up sources (which took them longer than an agent or LLM these days). Compressing that into a report might make it a review, but anyway. Great that you assembled a useful work tool here for some who need exactly that.