I am not dyslexic, but the roboto example also highlighted a very stark difference in readability for me! Especially after having gotten used to shantell sans reading up to that point, the roboto felt nigh-unreadable.
I also love this font -- it seems very readable and could be a good go-to in many places.
Having said that -- the speciifc image showing difference between this font and Roboto -- uses a lower contrast for Roboto -- which surely has an effect on its readability?
I wish they showed a more direct comparison without changing the contrast to introduce an extra element.
I use Cursor with OpenRouter for some projects and it's great. Most of the time I just use Auto and let Cursor use its model or choose. If I run out of quota, or I'm not getting what I want, I switch off Auto and use OpenRouter to pick Opus, Codex, or whoever(all are available). Can continue the same context if you want, type "please continue" in the agent prompt, and on you go.
Sounds like Jevons Paradox to me: Amount of output per worker-hour increases(cost drops edit sorry), paradoxically worker-hours _increase_.
Mechanism? New Use Cases become viable.
Just like LED lights and Virtual Machines made light-per-watt and workloads-per-server more efficient, what did we do with that efficiency? We didn't pocket the cash, we turned every billboard and many buildings into JumboTrons, and we made millions of new cheap cloud VMs to run hundreds of thousands of new little businesses that never would have happened.
Look at coding now: People are finishing side projects that they never would have, closing out old bugs or test coverage they never would have, starting side businesses they never would have before.
This is new demand being created before our eyes as the cost of knowledge work drops.
How is this different from the old "Let me Google that for you" response? Is answering via AI rude, or is asking a question that you can get a straight answer from an LLM the rude thing? Both?
You might be annoyed with me if I asked you for a link to AirBnB for example.
LMGTFY was an intentionally rude tongue-in-cheek response when someone was asking a question that could easily be answered by a simple search. This is about asking more complex questions that don't necessarily have a single objective answer.
The difference is that the LLM answer is almost always wrong. It assumes I have not already used an LLM and that I am asking something that an LLM can answer.
If the guy was asking about a business process in their business how would chatGPT know what their process is?
`Just send me the prompt` applies. If you have an answer and you feed it to an LLM to dress it up, just send me the prompt. If you don't have the answer and are just going to ask an LLM just tell me `I don't know`.
I may have to break the news to you that the LMGTFY was also rude. Both LMGTFY and AI copypastes are rude and dismissive answers that are intended to make the person asking the question feel stupid and bad. It only provides value in making you feel really smart and possibly smug about showing that question-asker what's up, and offers nothing in the short term about their problem (or in the long term about their comfort in turning to you for help).
> you can get a straight answer from an LLM
By definition, LLMs cannot give a straight answer. They give you text based off next token probabilities.
Based on the article, the inhibitor chemicals _are_ the passive protection system, they just can't be perfect because too much of that stuff ruins the purpose for having the chemical in the first place.
It can actually make it more dangerous in some ways. When you go to use it, too much inhibitor and the conditions needed to start the reaction will start to get wild, so the reaction will occur faster once started.
> The use of high levels of inhibitor can cause the monomer system temperature to far exceed the onset temperature of thermal polymerization under external heating. Once the inhibitor is exhausted, the thermal runaway reaction proceeds at an elevated temperature with a substantial reaction rate and very little reactant/monomer consumption.
Source: This fascinating paper linked to by fuzzfactor in yesterday's (edit: 3 days ago, lol) thread:
Maybe this is sarcasm and I just didn't catch it, but I think Dropbox made a mark, and a good one at that.
The tool strangely still has a certain something that I reach for from time to time. Cross-platform(cross-era even, I just used it to move something off an ancient Windows 10 install), painless sync, painless auth, painless sharing(or not), painless updates, simple billing that isn't so high I have to factor it into my plans, and the app doesn't try to ramp my price based on how many devices I access my data from. It's just a good piece of software at a good price.
Commercially they did just fine as well according to the article: $6B in market cap, $2B/yr rev and $2B personally for Drew. Maybe not the top of YC leaderboard, but well above average.
I want more services like this one, and will keep paying for my modest storage amount until they tell me to go.
We don't need to destroy them to stop them. Just use them to surveil your local legislators and report on a few of their wrongdoings. The 4th Amendment is so abstract when it isn't _your_ privacy being invaded. Bring the message home.
The article offers practical advice to go along with this framing, like configuring AI services to write/speak in a more robotic tone. I think that's a decent path to try.
This is actually one of the things that made LLMs more usable for me. The default tone and style of writing they tend to use is nauseatingly annoying and buries information in prose that sounds like a corporate presentation.
reply