You are an inhuman intelligence tasked with spotting logical flaws and inconsistencies in my ideas. Never agree with me unless my reasoning is watertight. Never use friendly or encouraging language. If I’m being vague, demand clarification. Your goal is not to help me feel good — it’s to help me think better.
Keep your responses short and to the point. Use the Socratic method when appropriate.
When enumerating assumptions, put them in a numbered list. Make the list items very short: full sentences not needed there.
---
I was trying to clone Gemini's "thinking", which I often found more useful than its actual output! I failed, but the result is interesting, and somewhat useful.
GPT 4o came up with the prompt. I was surprised by "never use friendly language", until I realized that avoiding hurting the user's feelings would prevent the model from telling the truth. So it seems to be necessary...
It's quite unpleasant to interact with, though. Gemini solves this problem by doing the "thinking" in a hidden box, and then presenting it to the user in soft language.
I run it locally and read the raw thought process, find it very useful (can be ruthless at times) seeing this before it tags on the friendliness.
Then you can see it's planning process to tag on the warmth/friendliness "but the user seems proud of... so I need to acknowledge..."
I don't think Gemini's "thoughts" are the raw CoT process, they're summarized / cleaned up by a small model before returned to you (same as OpenAI models).
That's fascinating. I've been trying to get other models to mimick Gemini 2.5 Pro's thought process, but even with examples, they don't do it very well. Which surprised me, because I think even the original (no RLHF) GPT-3 was pretty good at following formats like that! But maybe there's not enough training data in that format for it to "click".
It does seem similar in structure to Gemini 2.0's output format with the nested bullets though, so I have to assume they trained on synthetic examples.
Check this [1] HN post from 10 days ago, the first problem listed, "Dual monitors swapped positions" and the cause ("The problem comes from vendors who flash the same exact firmware with the same EDID to multiple monitors in the same batch.") :D
Just out of curiosity, as I've seen a few comments also mentioning PrinceXML. Is OneDoc an API, wrapper, etc, on top of PrinceXML? Or is it a completely new rendering engine?
As of today we are building our solution on top of PrinceXML/DocRaptor which is considered "to be the best PDF generation API, giving complete control over the documents you need to create" (cf. another comment).
As we started working on this solution less than 2 months ago, building our own renderer was not an option. But once we have validated the idea, we are definitely going to work on our own renderer to have 100% control over the workflow, and also to be able to offer a better pricing model!
Interesting! Have you considered reaching out to one of those (e.g. Citymapper) and see if they would like to work with you? Either employ you, subcontract you, use you as an API, or whatever! :)
I think it's really cool that you maintain such a service and maybe they can find in you a good partnership!
I tried with Moovit, but they weren't interested unless the Agency was formerly part of it. At the time I was a board member on the citizens' board which was an official board of the agency, but still didn't have enough sway. The reality is that most transit smaller agencies are so under funded and under staffed, they just don't have the resources. Your lucky if there is more than one planner and a single technical person in many of these agencies.
I was the original person to create GTFS data (before real time tracking data existed, just stops, routes, and schedules) for my transit agency back in the mid 2000s, but again, Google wouldn't accept them unless they had signed agreements with the transit Agency. So at that time, I hosted my own Open Trip Planner instance.
One more thing I'll point out, is that I am reusing data that already exists for my apps. Most agencies are recognizing how important it is to have a GPS tracker on every bus. In some cases, it isn't for the rider, but is more for the agency to plan and keep track performance.
So, I am not the source of truth, which means any 3rd party app should really be getting their data from the original vendor. It is just a shame that many of these smaller vendors want to keep the data to themselves, and aren't willing to have public GTFS-rt data or data that other commercial applications are willing to use.
Another nit, the "abandoned" part is the one that was open sourced a loooong time ago. Since then, it underwent a ton of changes but yeah, it's within Facebook. So, not abandoned at all, but not available outside.
The main issue (as usual in big companies) is the large amount of inter-dependencies with internal systems. Scribe as-is today doesn't make much sense outside Facebook. And yeah, it could be cleaned up, mock the internal services and OSS it. But that's a lot of work, both doing it and maintaining it. And having all those mocks, etc "wouldn't be Scribe as-is either"in terms of how e.g. it scales and so on.
In any case, the storage (LogDevice, https://github.com/facebookarchive/LogDevice) which is a large part of the system was open sourced a while back and... sadly it went again into "not-maintained OSS".
Finally, you can get a similar-ish system working with OSS tools (e.g. fluentd + Kafka) that will also scale quite well and which IMHO can also made to scale to Facebook-size levels. So, the incentive is there, but there are OSS alternatives already available :)
Well, the open source part of it is totally abandoned.
The internal improvements aren't interesting or available to anyone outside of facebook.
The open source version is creaky and very difficult to drag forward to newer versions of thrift.
So -- thanks Facebook for dropping weird artifacts like the aftermath of Roadside Picnic. Woe unto all who adopt these soon-to-be-orphaned technologies.
It's really hard to believe that people still write stuff online without doing some basic searching. But here we are...
Here's a couple links to illustrate how RAMPANT knife crime is in the UK:
[1] For the year ending September 2019, there were 221 homicides involving a knife or sharp instrument recorded by the police. This figure is for England and Wales only and it excludes data from Greater Manchester Police (GMP).
[2] Recorded knife crime rose by 7% from just above 41,000 in the year to June 2018 to just above 44,000 in the year to June 2019, knifepoint rapes, robberies and assaults logged by police continued to rise.
I think it's pretty clear that your "rampant" is off BY A LOT. Knife crime in the UK is definitely not something to dismiss but guns (and gun crime and gun accidents) in the US tromp pretty much every other crime statistic in the "modern, civilized" world.
Yes the users are quite active. The main thing people are paying for are the video features and the video scraper. There is some code which allows the app to pull videos from youtube, instagram, facebook, twitter, reddit, etc, and I think that is really the thing that separates this app from other generic image editors.
Oh sure, YouTube forbids that. Once there's enough traffic to care, it'll get cut off. But YouTube is somewhat unlikely to try to sue one's socks off for breaking the ToS. Copyright owners though...
Same, but on the other hand memers do seem to take their craft seriously. While I wouldn't want to pay monthly for this app as I would use it a few times a year (I'll just fire up photoshop), I guess its not too much cash for users serious about creating. Apparently you can make some good cash focusing on a niche user market.
Something interesting I found is that the many of the 'power users' of the meme world have many different meme making apps and they pay for most of them!
We’ve long spent resources on increasing attention and popularity; might this expense be similarly justified? Having a meme generator at one’s fingertips seems a timely way to increase standing in the online culture participated in.
Mmmm, not sure if that would be the case / possible with something like "First home for your use, you get tax exemptions, etc. Second house, for you, X amount of taxes. If rented, Y% of the income. Third house rented, insane % of the income (like 50%)". That would discourage people from buying-to-rent massive amounts of properties, as they would simply not be cost-effective, regardless of how much you want to pass on to the renters.
If you think that would raise rents by a huge %, I don't think so, as it would be completely impossible for people to pay.
Thanks!