184 karma · joined September 26, 2022
https://convergence-frontier.com
Whether or not a general purpose foundation model for coding is trained on more backend or frontend code is largely irrelevant in this specific context.
It's made me realize that objects are much further from the boundaries of my car when backing into a spot parallel parking. I would never think to get so close to another car if I had to only rely on my own senses.
With that said, I realize there's a significant number of people that are even poorer estimators of these distances than myself. I.e. those that won't pass through two cars even though to me it's obvious that they could easily pass.
I have to imagine a big part of this has to do with risk assessment and lack of risk-free practice opportunity IRL. Nobody is seeing how far they can push or train themselves in this regard when the consequences are to scratch up your car and others' cars. With the birdseye view I can actually do that now!
Random callout: the copy in your app store preview images would benefit from some proof reading. Example: "WE dont just store thoughts, but makes sense of them" should likely be "ThoughtCatcher doesn't just store thoughts, it makes sense of them". My 2 cents is to also rework "Capture your mind" as it's a little awkward. Maybe "Organize your thoughts", "Supercharge your thoughts", or something along those lines.
The key seems to be in curating your application's evaluation set.
I'll beginning to integrate it into my user-facing application for language learners soon: www.abal.ai
For better or worse, you won't find correctness on any business' income statement. Sure, it's a latent variable, but so is efficiency.
You'll of course be free to use them optionally in your free time and on personal projects. It won't be the case at your place of employment.
I will mark my calendar!
Which models have you tried to date? Can you come up with a top 3 ranking among popular models based on your definition of value?
What can be said about the ability of an LLM to translate your thinking represented in natural language to working code at rates exceeding 5-10x your typing speed?
Mark my words: Every single business that has a need for SWEs will obligate their SWEs to use AI coding assistants by the end of 2026, if not by the end of 2025. It will not be optional like it is today. Now is the time you should be exploring which models are better at "thinking" than others, and discerning which thinking you should be doing vs. which thinking you can leave up to ever-advancing LLMs.
This is especially true for conventional SaaS that historically required one or more developers to read and internalize documentation for a service's API in order to make any use of it.
What are the odds that this developer fully understands how to squeeze every bit of value of out this expensive SaaS product, for which the company is paying monthly/yearly?
What if that developer jumps ship? Or the provider releases a major version bump with a significantly different API, but features that the business absolutely needs to integrate in order to stay ahead?
As much as it might seem hyped up at the moment, MPC and agent-based tool calling paired with CoT can and will make things orders of magnitude more efficient.
No business in its right mind will accept anything less than the total integration of MCP and agents across their entire stack by the end of this year.
No sane organization will accept anything less than forcing SWEs to use these tools moving forward. The details might differ and some will delay because "IP", but they'll all converge very quickly.
Not to say there's no joy in doing it the old-fashioned way on your own time :)
I've seen the light!
Most recently, gemini 2.5 pro makes the agentic workflow usable, and how!
also, will I be able to play pool?
What am I doing to fix this long term?
I'm currently building an optimization mechanism to ensure the highest quality translations for any given language pair.
I really appreciate your detailed feedback. Do let me know if anything else comes to mind.
Systems like IPA (https://en.wikipedia.org/wiki/International_Phonetic_Alphabe...) don't make pronunciation any easier to get right for a majority of - if not all - language learners. ABAL solves this by showing phonetic hints that are written in the user's native language script and are meant to be pronounced in the native language's voice.
The name All Books, All Languages is intentionally quixotic, as it serves to set the ambitious and unreachable goal of hosting all of humanity's written works while making them available to all language learners and readers across the globe!
Today, it hosts a number of generated short stories categorized by subject and CEFR level. I'm currently working on supporting (private) self-uploaded files, and works in the public domain. Work is also going into adding support for more languages, while evaluating and optimizing for cross-language translations accuracy.
It currently supports 56 languages (including Klingon and Latin!) - which comes out to 3080 permutations - or 3080 unique user profiles as I like to think of it. Both the marketing and application sites are fully internationalized - even response messages in 4xx responses use i18n. Quite a bit of work went into automating the i18n diff generation system to make sure I can release UI/UX changes quickly.
I hope you'll find it useful if you're learning a new language or need to practice one you already know. If you're not learning a new language, but wish you could read classics in a language you're fluent in, stay tuned!
While it's currently decent for the purpose of language learning, I do need to fix consistency across longer stories because a purely sequential generation suffers from consistency in plot details.
The idea was always to offer real books in the public domian, but I decided to release it with generated stories so I could begin gather feedback on UI/UX.
Thanks for calling that out! I'll look into it.