5,827 karma · joined November 21, 2016
We do get caught up in like "I shouldn't have said x or y" when the reality is if you make other people feel nice they will want to have you around. If you say the perfect words but overall make them feel bad they won't want you around.
To confidently interrupt I would want to assert that the user has been speaking for > N time. You could do other things like parse a streaming transcription for keywords but generally it feels like bad UX to me to cut output the second input is detected. Letting the user talk for 1-2s gives a much stronger signal and it isn’t too weird for someone to keep talking for 1.5s after you start.
1-2s replies feel natural and like you pointed out pausing for 2-3s mid sentence is super normal.
Curious if you thought their approach was necessary, it seemed like a ton of complexity to reduce one of the faster parts of a voice AI setup. Having a fast model and accurate VAD seems way more important than fine tuning WebRTC transit times.
If tech companies invested 10% of what they have in AI assisted coding tools into AI assisted legal tools, they would be able to do those steps easily.
It is definitely coming.
How much does it cost to send a billion people an hour of video every day? Almost all of the resources tech uses is for pointless or even negative things.
What % of compute/bandwidth do you think is used for "real value"? I would guess it is well below 1%.
I always wondered about this statement, like we are generally salaried and there is so many variables that affect how I spend my "time". None of us are machines that can do X work per day and our managers get to slice it as they see fit. Pull a dev off a project they love and throw them onto something they hate and suddenly X is diminished greatly.
I would almost predict that reshaping our workflow to be: "prompt, wait, approve changes." results in losses because it is such a mentally tiring workflow and drills into our brains the desire for the LLM to "just fix it". It is the next level of just moving tickets to completed all day.
That is no longer a helpful tool... it costs like ~15% of an actual dev.
Even if it is helping, is it actually... making things better or building anything truly important? The issue seems way too nuanced to spend $2k/mo. Not to mention the entire tech industry floats on hype and imaginary goal posts so now what? Devs can hurdle towards those faster and more mindlessly?
Acknowledge it is normal? Attempt to buy deeper into the delusion ("Yeah my work is awesome and unique!"). Use stimulants to force enthusiastic days every once in awhile?
Even if we assume those giant leaps of faith are true, it still means "I" go away when I die. My soul will only remember "me" as a brief torture where I was forced into a human shell and had to endure being "me" before being released once more to be a higher being. All my struggles, battles, self-improvement, etc. will be meaningless and my kids will be a cruel trick I played that imprisoned other souls.
Which is really heartening to me, because decreasing the temperature of the ocean seems daunting, but not dragging giant nets through the ocean nonstop seems pretty straightforward.
I always liked the fence story better though.
Reading through the list mostly made me feel sad. You can't help but interpret these through the modern lens of AI assisted coding. Then you wonder if learning and following (some) of these for the last 20 years is going to make you a janitor for a bunch of AI slop, or force you into a coding style where these rules are meaningless, or make you entirely irrelevant.
One hard hit in the head can literally change your personality entirely, then you have alzheimers and all the other degenerative brain diseases that will erase "you". Even if you avoid all that "you" will be wildly different every 15-20 years.
Christianity gets through this by saying you will return to your prime. That just seems kind of childish to me though, like "yeah when you die you and all your friends and family are gonna be 25 and you live in paradise together forever".
How do you resolve the idea of an eternal consciousness with the very real and common occurence of people losing their consciousness while they are still alive?
There is a class next door to my office. An old woman is teaching ~20 people how to be insurance agents with a slide show. It seems like a two week course with a certificate at the end.
They don't seem worried that the slideshow could be pasted into an LLMs context window and outperform all of them on the test in 5 seconds and are diligently taking notes.
Like years of manually studying, fixing and reviewing code is experience that only pre ~2020 devs will have.
The intuitive/tacit knowledge that lets you look at code and "feel" that something is off with it cannot really be gained when using Claude Code, it takes just 1000s of hours of tinkering.
It will suck if the job shifts to reviewing and owning whatever an LLM spits out, but I don't really know how effective new juniors are going to be.
Obviously it would be a dystopian nightmare to have everyone yelling inputs into their phone on the sidewalk, but at certain times it would be extremely useful (while driving, etc.). It allows for a crazy level of accessibility, and sometimes I just want to not stare at a screen or type anymore.
With that in mind I made https://veform.co. Still a million miles from the dream but it has working demos and a playground to code different form conversations in.
Are there any complications with that or is it kind of on me to not confuse the other peer by sending unexpected formats?
I always wondered if there were nutrients we can’t measure though, because collagen is typically made from skin or bones whereas whey protein is dairy. Even though both have similar “nutrition facts” maybe there are some unmeasurable differences. Both are highly processed though so who knows.
Hypothetically, you have a simple slice out of bounds error because a function is getting an empty string so it does something like: `""[5]`.
Opus will add a bunch of length & nil checks to "fix" this, but the actual issue is the string should never be empty. The nil checks are just papering over a deeper issue, like you probably need a schema level check for minimum string length.
At that point do you just tell it like "no delete all that, the string should never be empty" and let it figure that out, or do I basically need to pseudo code "add a check for empty strings to this file on line 145", or do I just YOLO and know the issue is gone now so it is no longer my problem?
My bigger point is how does an LLM know that this seemingly small problem is indicative of some larger failure, like lets say this string is a `user.username` which means users can set their name to empty which means an entire migration is probably necessary. All the AI is going to do is smoosh the error messages and kick the can.
I struggle to believe that a ton of seemingly intelligent software engineers are too dumb to figure out how to use Claude code to get reliable results, it seems much more likely to me that it can do well at isolated tasks or new projects but fails when pointed at large complex code bases because it just... is a token predictor lol.
But yeah spinning up a green fields project in an extensively solved area (ledgers) is going to be something an AI shines at.
It isn't like we don't use this stuff also, I ask Cursor to do things 20x a day and it does something I don't like 50% of the time. Even things like pasting an error message it struggles with. How do I reconcile my actual daily experience with hype messages I see online?
Like I wouldn't expect any other frontend router library to decorate links for me in a certain way.
I did notice that forward doesn't seem to work. I.E. If I click into a page, it zooms in, press back it zooms back out, press forward it flickers the url but doesn't have normal forward behavior.
I also don't know if you want to support `open in new tab` but that would be a hard req for many people.
E.G. if your name is `Donold` (pronounced like Donald) there is not a transcription model in existence that will transcribe your name correctly. That means forget inputting your name or email ever, it will never output it correctly.
Combine that with any subtleties of speech you have, or industry jargon you frequently use and you will have a much more useful tool.
We have a ton of options for "predict the most common word that matches this audio data" but I haven't found any "predict MY most common word" setups.
My GF, daughter and me all play Stardew Valley but we play it wildly differently. It is a farming/relationship simulator for them and some kind of capitalist min/max farming and mining simulation for me.
But yes, the Sims 3 and the 700 add ons are all heavily in their rotation, they make me look like a gaming amateur if you go by hours logged.