Same? Not quite as good as that. But google’s Gemma 3 27B is highly similar to their last Flash model. The latest Qwen3 variants are very good, to my need at least they are the best open coders, but really— here’s the thing:
There’s so many varieties, specialized to different tasks or simply different in performance.
Maybe we’ll get to a one-size fits all at some point, but for now trying out a few can pay off. It also starts to build a better sense of the ecosystem as a whole.
For running them: if you have an Nvidia GPU w/ 8GB of vram you’re probably able to run a bunch— quantized. It gets a bit esoteric when you start getting into quantization varieties but generally speaking you should find out the sort of integer & float math your gpu has optimized support for and then choose the largest quantized model that corresponds to support and still fits in vram. Most often that’s what will perform the best in both speed and quality, unless you need to run more than 1 model at a time.
To give you a reference point on model choice, performance, gpu, etc: one of my systems runs with an nvidia 4080 w/ 16GB VRAM. Using Qwen 3 Coder 30B, heavily quantized, I can get about 60 tokens per second.
I get tolerable performance out of a quantized gpt-oss 20b on an old RTX3050 I have kicking around (I want to say 20-30 tokens/s, or faster when cache is effective). It's appreciably faster on the 4060. It's not quite ideal for more interactive agentic coding on the 3050, but approaching it, and fitting nicely as a "coding in the background while I fiddle on something else" territory.
Yeah, tokens per second can very much influence the work style and therefore mindset a person should bring to usage. You can also build on the results of a faster but less than SOTA class model in different ways. I can let a coding tuned 7-12b model “sketch” some things at higher speed, or even a variety of things, and I can review real time, and pass off to a slower more capable model to say “this is structural sound, or at least the right framing, tighten it all up in the following ways…” and run in the background.
Well theres an open source GPT model you can run locally. I dont think running models locally is all that cheap considering top of the line GPUs used to be $300 now you are lucky if you get the best GPU for under $2000. The better models require a lot more VRAM. Macs can run them pretty decently but now you are spending $5000 plus you could have just bought a rig with a 5090 with mediocre desktop ram because Sam Altman has ruined the RAM pricing market.
The run at home was in the context of $2k/mo. At that price you can get your money back on self-hosted hardware at a much more reasonable pace compared to 20/mo (or even 200).
I got some decent mileage out of aider and Gemma 27B. The one shot output was a little less good, but I don’t have to worry about paying per token or hitting plan limits so I felt more free to let it devise a plan, run it in a loop, etc.
Not having to worry about token limits is surprisingly cognitively freeing. I don’t have to worry about having a perfect prompt.
No it’s like someone owning a Ferrari and looking down on someone who drives a Corolla. Or that’s how they see it, anyway. Plus there’s the annoyance with interoperability: it’s not just about status, it’s about all your iMessage group chats that don’t play nice with android
I'm confused, I thought we were talking about people who are installing and running openclaw. You're right, if this is now a thread about teenage dating habits, I'm out.
iMessage lock in is a huge thing. When it was new and was still e2ee I ended up buying iPhones for everyone I regularly messaged.
These days it is insecure however because they backdoored the e2ee and kept it backdoored for the FBI, so now Signal is the only messenger I am reachable on.
Blue bubble snobbery is presently a mark of ignorance more than anything else.
I agree that it’s stupid to judge people for it, but you do have to admit that especially with not all people having RCS, the feature set of SMS and MMS that you have to deal with when not using iMessage is pretty barbaric. From the potato-quality videos (ironically, I recall QuickTime was heavily involved in that spec, lol) to the asinine way Apple lets you apply a reaction and then sends it as a verbose text… From an iPhone user’s point of view, a “green bubble” means “this conversation will work like it’s 2003.”
Yes, I know 99.999% of Android users are on WhatsApp (or WeChat, Line, or Telegram depending on cultural background) but at least half of iPhone users aren’t on those, so we still have to keep using Messages for a lot of people.
Removing the skill does remove a level of indirection.
It's a difference of "choose whether or not to make use of a skill that would THEN attempt to find what you need in the docs" vs. "here's a list of everything in the docs that you might need."
I think if you read it, their agents did invoke the skills and they did find ways to increase the agents' use of skills quite a bit. But the new approach works 100% of the time as opposed to 79% of the time, which is a big deal. Skills might be working OK for you at that 79% level and for your particular codebase/tool set, that doesn't negate anything they've written here.
As far as I can tell, the nearest thing to a stated goal or mission is on their “About” page:
Our main features are:
* ReactOS is able to run Windows software
* ReactOS is able to run Windows drivers
* ReactOS looks-like Windows
* ReactOS is free and open source
Building a replica of an old OS is a fun project, but if there was a purpose for it besides having an "is able to" replica, it would attract more people.
in the real world, most people use windows. most software that those people use is written for windows. if it can run windows exes out of the box, whilst not phoning home to microsoft, it becomes an attractive proposition. i want to get off windows but i dont want the headache of linux; to me its the only hope
Sure, but Windows has moved a long ways since the version that they're attempting to replicate. And again, their bar for success "is able to run Windows programs" is not actually high enough to achieve a practical Windows replacement, even if going back to Windows 95 is all we wanted.
It's interesting you mention Linux being a headache — it is, but there is an order of magnitude more people working full-time on just the Linux desktop experience than have ever even tried running ReactOS. That ratio would have to flip before the latter has a hope of being a useful Windows replacement. We’re much more likely to see Wine able to run 100% of Windows before ReactOS gets there.
> And again, their bar for success "is able to run Windows programs" is not actually high enough to achieve a practical Windows replacement, even if going back to Windows 95 is all we wanted.
how no?
> It's interesting you mention Linux being a headache — it is, but there is an order of magnitude more people working full-time on just the Linux desktop experience than have ever even tried running ReactOS
and hows that going these days? still a nightmare? basic functionality introduced maybe 50 years ago now, and the linux world is still working out the kinks with GUIs, probably part of the reason TUIs are becoming popular.
Minnesota has a better system. You fill in a paper ballot using a pen, and the paper ballot gets optically scanned.
Besides avoiding any issues (real or imagined) with touchscreens, it makes it extremely cheap to stand up more polling places with more booths, since only one tabulator is needed; the booths themselves can just be little standing tables with privacy protectors.
>Minnesota has a better system. You fill in a paper ballot using a pen, and the paper ballot gets optically scanned.
>Besides avoiding any issues (real or imagined) with touchscreens,
Wait... I don't think these are the complaints being made against internet voting at all. The problem is with a computer counting and reporting it, right? Centralized, less transparent, etc.
I dont view writing my vote on paper and scanning it to be paper voting if it's just immediately fed into a computer.
> I dont view writing my vote on paper and scanning it to be paper voting if it's just immediately fed into a computer.
The paper ballots are retained for recounts, and most places with this system automatically recount a random subset of the paper ballots to ensure it matches the computer totals. This guards against both shenanigans and mistakes. For security the scanning machines are not networked! A person carries around a little SD card (not USB as it's too hackable) to collect the totals.
The paper ballot with in-precinct immediate scanning system is the best system I've seen. It reports results quickly and leaves a full paper trail for recounts and accountability.
They are USB on the machines we use. That said, that’s not a concern to me.
The machine also prints a paper tally that goes with it to verify. We used sealed bags so they can’t be messed with in transit. They tabulate the results and compare it to the total from the tape. Personally, I wish there was a hash of the results that would make it simple to say “yep, that’s the same” but practically that’s not necessary.
A second copy of the receipt goes back separately with the paper ballots. Same sealing and chain of custody handoffs.
I like the electronic ballot marking device. I can understand the argument that they’re not worth the cost, though.
This was common in Texas, but becomes challenging when one polling place serves voters that might have different elections to vote for - say, at a polling place on the line between two school districts or something like that. You can't just print one sheet of paper, and it to everyone, and call it a day. Toss in a few different jurisdictions that don't directly overlay each other, and the number of combinations become nontrivial.
(the machines used in Texas vary by county, in my county we use Hart InterCivic machines that are touchscreen but produce a paper trail - honestly I think it works well)
This really is the best way to do it. Scantron gives fast results and you get a paper physical record which shows the actual ballot exactly as it was presented to the voter along with what their vote was.
<devilsAdvocate>How many people spend time making their selections on the computer, then compare every single selection on the print out? Deniers could say the computer randomly prints votes to skew in certain candidate/party direction knowing not everyone would catch it.</devilsAdvocate>
all it would take is one person saying their printed ballot does not match their specific selection, and the whole thing would become chaos.
Same but different issues. Now you have to know that the dots were filled in correctly to be readable. Having someone make an obvious attempt at selection but not readable by the reader is also problematic. No reason to not count their vote. You may laugh about not being able to do it correctly, but it happens.
Only if the scantron shows that each position on the ballot was counted and the voter is not allowed to leave until the person monitoring the scan confirms with the voter their ballot was scanned would this give confidence. Any issues with the scan, and the voter is allowed to correct the issue. There should never be an issue of reading the ballot by the scanner as an acceptable outcome.
of course, all of this is assuming in person voting only
Checking each ballot for completeness sounds like a good improvement to the system. Right now people are just expected to mark carefully and double-check their work before feeding their ballot into the machine and request a new ballot if they mess up.
It might slow things down a little bit, but making sure that the machine can detect a vote for each race/question (even if it's just "Abstain") would make sure people didn't forget to fill out something and help prevent the fill-in-the-bubble equivalent of hanging chads.
I like the idea that "abstain" should be an option for each position on the ballot to remove the ambiguity of it just being skipped mistakenly. Require every position on the ballot to need a response from the voter regardless. That would definitely simplify the tally process even if it does require the voter to go back to fill in additional spots. Better to be right on even if it takes 30 more seconds.
We agree. Don't use computers. Scantron is only there to get a fast count for the news agencies. Manual counting of physical paper ballots would still be done anyway.
To manually count by hand every ballot would mean not finding out a complete tally well until after Jan 20. When election day and inauguration day was selected, the number of ballots to count were a mere fraction of today's count.
Manually counting votes is so error prone that I'd have less confidence in it than a scantron type of ballot. At this point, I'm more in favor of giving each voter a ball/bead/chip to drop into a bucket for each position on the ballot. After checking in, you go to each position to receive your one token. If you don't visit a position, you do not get a token to pass to someone else. Tallying the votes could be as quick as weighing the bucket as the weight of the bucket/token will be known. Each election can change size/weight/color of tokens to be unique. If the weights total an irrational weight, it would be deemed suspect and a hand sort of the tokens can be done to find the odd token.
Hand counts are kind of obnoxious but they can't be beat for transparency. There's no reason it has to be done at once either. Ideally people would be able to vote over several days and counting can start right away.
Balls/tokens aren't a bad idea either though, but it sounds like people pocketing a ball/token would force a manual count even if they kept them since the total weight of all buckets combined would be off. I'd also worry about people bringing in heavier or lighter balls/tokens but the bigger risk would be poll workers handing out heavier or lighter balls/tokens to specific people (or types of people) because it'd be easier to make sure the weights would add up in the end.
Maybe we could force everyone to vote at every position (which should have an abstain option) then have the machine check the weight of every ball/token as it was inserted, and verify that one but only one was inserted, before it fell into the selected bucket?
To me, hand counts are beyond obnoxious. How many times does each ballot need to be counted? Just once? Someone with an agenda could cause havoc. Twice? Three times? Majority wins? How many times would non-unanimous count be allowed before the person making the odd result be dismissed/replaced? I can't remember the hanging chad debacle process, but I do seem to remember one person looking at it before handing it to the next person for confirmation.
I like the idea of placing the token into a verifier to validate authenticity before dropping into the bucket. Similar to a coin sorter where invalid tokens get rejected to a separate bin with a light and siren to ID the person trying to cheat. These could get expensive as you'd need one per candidate per position on the ballot.
Transparency comes much more from repeatable results than manual process. You run the same stack of 1,000 ballots through 2 optical scanners, they will give the same result unless one is busted (in which case do it with 3 or 4). This takes very little time and is reliable. Do it by hand and you are guaranteed to get a different result almost every time, and it will take forever.
reply