Windows 7 was the last great Windows (it took the best of NT4 and 2000 and put it in a consumer package), and like you, I've used them all (Bob doesn't count). I lived with 10 for a long time, and Microsoft was trying to shoe-horn a lot of the nonsense you see in 11 into it. Ads on the Start Menu, and the dock.
But they've gone full-tilt Bozo with 11. The ability to deliver such an experience to their advertisers and marketing resellers was the whole reason for its existence. That is, 11 is about what Microsoft can get from its users, not about letting them use their own computer. It is no longer a suitable personal computing OS.
Warts and usability issues are present in Tahoe, and I wish Apple hadn't made the choices they've made on the UX, but Tahoe remains closer to being something for the user than Windows is.
Linux is generally for the sake of operating a computer (I didn't forget the user, but Linux--sweeps hand--makes the assumption that it's users are developers). That's what's so surprising about the age verification push in some Linux distributions (and their attendant bans for disagreement in the mailing lists or on GitHub).
For Star Wars, they retconned it to mean he found the shortest possible route through dangerous space, so even for Han Solo's quote, it's still distance.
Please protect your supply-chain. Python projects in the wild, especially as relates to agents, give me the willies these days with the GitHub Actions compromise that's been rolling through the OpenClaw dependencies. (Last one I saw was Telnyx.)
How does this blend with something like wedow/ticket, or beads? Those are "short term" in-repo task lists that help shape context. I've been thinking that adding a good long-term memory was the next step. I'll have to kick the tires tomorrow.
Thanks so much for the comment! Yes, I initially had LiteLLM gateway and backed it out after the security issues.
Memory is becoming a crucial part of agents and there isn't a one size fits all solution - unfortunately. I found I wanted to replicate things like SOUL.md but in a way that could "evolve", so this project has "frames" which is just a collection of tagged / filtered memory blocks which can be used to identify specific concepts.
It also allows the agent to use "outcomes" to calibrate it's own memory blocks, which is really powerful.
Anyway I think "cognitive loops" are going to become more relevant, and will be an art. Thanks again... let me know how you get on!!
A number of years ago I wanted to drop a webhook when a call came in on VoIP.ms but couldn't find any way to do it natively.
Ended up sticking a twilio endpoint in the ring group with a "press 1 to accept this call" message so it wouldn't eat the call, then was able to fire an http request with the call details.
It worked well, although I admit I was a little annoyed I couldn't do it directly with VoIP.ms.
Because they exclusively used a model that was about as big as the original GPT-2.
Which, I mean, fair enough within these constraints, but it's cited like it's a universal law.
Really all that can be taken away from the study is "we trained a very small model on data generated from it in a particular way, and this was eventually harmful for the model."
Also note that models are nowadays trained on massively self-generated data (task RL post-training) and it seems to significantly improve their performance.
Clojure is the most token efficient of the major languages. It's a pleasure to work in. The sample code base that Claude knows skews toward higher quality code than the large volumes of middling JS or Python (this is not about the language but about what gets shared in the open for training data).
You want binaries... Babashka.
CLI... sure.
Access to the entire Java ecosystem... also yes.
The one other virtue is that when I use Clojure to demonstrate some feature, no one at work is tempted to deploy it as is until they rewrite it for Node.
Perhaps I'm naive, but why would any military, anywhere on Earth, use a service connected over Internet 1 as part of their logistics and planning chain?
Or do these contracts already stipulate operation on custom hardened network connections?
U.S. military hosts the single largest intranet in the world. It parallels the architecture of the Internet, but is entirely self-contained. Contracting with SaaS tends to revolve around RFP's in which you end up establishing an enclave within the Intranet, or (during the time when I was there and they were still flirting with the whole cloud thing), setting up a partitioned, high assurance, tunneled "over the Internet" into the Intranet DMZ. There were also many growing pains where different branches of the military realized SaaS wasn't really compatible with their operational domain. My "crowning moment" of finally becoming a true blooded Dev in the eyes of my coworkers was when a 6+ month project I got to the point of finalizing with a negotiated agreement between service provider and the agency in question got shit canned due to logistical incompatibility of the desired business model of the service provider, and the war fighting realities of the context in which it would be used. It was... A truly bittersweet feeling killing that metaphorical baby. It did teach me Rule #1 of Real Software Architecture though. Always check the legal copy/fine print before implementing something with the technical capability to work. Better to spend a month reconciling the service provider's expectations of how their operating relationship is going to look like first so you can suss out whether or not the development time to integrate this dependency is even worth it. It's an absolutely brutal lesson, that only tends to sink in successfully once you realize that your last 6 months of gestating a systemic augmentation was essentially a government subsidized miscarriage. It marks you. Or at least it did me. In the case of Anthropic, it sounds to me like someone went in, deemed the redlines non-issues since they were actually probably evaluating the throughput amplification potential on the logistical front, then someone had the bright idea of " but what if we used it here, without checking the contract or RFP scope first. In any other Admin., it would have been a mea culpa, and a training experience for someone in vendor relations and government procurement. In this admin, well... Not so much. It won't be the first time; as I can personally attest, and it won't be the last by a long shot. It's what makes Government work especially complicated.
When I was in the military, they were very much against using technology and internet. Carrington was a thing, nukes/emp are a thing. There's no guarantee that technology would continue to work when the military is needed most.
I believe that must have changed a lot in the last 20 years for this to even be a discussion.
Someone on hackernews[0] told me that Anthropic models are accesible using amazon bedrock service which is available within AWS secret regions from my understanding which is Amazon's datacenters specifically designed for military.
Are you saying it's not acceptable for a woman to choose a female driver over a male driver for a sense of her own safety?
Deep breath in... There are two types of discrimination. Paraphrasing Thomas Sowell, let's call them Type I and Type II.
Type II discrimination is the evil awful kind we rightfully rail against. It is "treating people negatively, based on arbitrary aversions or animosities to individuals of a particular race or sex..."
Type I discrimination is of the broader sort; "an ability to discern differences in the qualities of people and things, choosing accordingly." We run our lives with this kind of discrimination: is this food safe to eat? is this activity safe to participate in? do I trust this person given what I know about them?
>> Ideally, Discrimination I, applied to people, would mean judging each person as an individual, regardless of what group that person is part of. But here, as in other contexts, the ideal is seldom found among human beings in the real world, even among people who espouse that ideal. If you are walking at night down a lonely street, and see up ahead a shadowy figure in an alley, do you judge that person as an individual or do you cross the street and pass on the other side? The shadowy figure in the alley could turn out to be a kindly neighbor, out walking his dog. But, when making such decisions, a mistake on your part could be costly, up to and including costing you your life. [1]
This kind of discrimination is what we're talking about. I'd venture that not only is it OK, it is necessary. In this case, men that have had no background check, and whose form of employment is as an Uber driver are more likely to harass women (or do worse) than a female driver. Allowing women to make a selection based on this likelihood means that female customers that are alone can make choices to still use the service while reducing the overall risk.
Mitigation of this risk in normal taxi services take the form of background checks, bonds, and a chain of responsibility running from employer to employee to customer. It places more risk on the employer deliberately. Uber deliberately chooses to avoid this risk and responsibility. That choice is baked into their business model. That means enabling this kind of discrimination from their customers is a required feature of the service.
[1] Discrimination and Disparities, by Thomas Sowell
> Allowing women to make a selection based on this likelihood means that female customers that are alone can make choices to still use the service while reducing the overall risk.
I'm failing to see how anything you say could be used as a guideline to pick between "good" discrimination and "bad" discrimination.
The major distinction you draw between "Type II" and "Type I" is the fact that one is fueled by "arbitrary aversion" which is not a particularly useful distinction.
What if I denied entry to black people from my bar because ""they commit more crimes"" and ""are more likely to break stuff"", is it morally ok? Why not?
My opinion is that no, it's not ok because the majority of people punished were never going to behave in an uncivil way.
The same logic can be easily applied to this situation. Are men more likely to behave sexually inappropriately (which ranges from verbal harassment to assault)? Sure.
Is it the majority? Hell no, it's nowhere close.
(Of course it's worth nothing that the "majority" does not necessarily have 50.01%, it's just an arbitrary line you can draw as long as you are consistent about it)
The point I took away is that since the normal methods of "ok discrimination" are not available and Uber refuses to do the needful on their behalf, women should be able to "use the big gun".
The reality is that if Uber rapes are an issue, and something like this is not allowed, women will just stop using it entirely.
Or special Uberpods will be developed where the driver is completely encased and the passenger has a "auto drive to police station" button.
If someone is presenting themselves to you in person for entry into your bar, you have far more information to make a judgement on than the color of their skin... so it is not the same.
In the case of a woman coming into contact with some driver and volunteering location information like her home address, she has little to no information to make that judgement. Providing her just that bit of information, and allowing her to discriminate based on it, makes her safer. Ideally, she'd have way more information than just whether the driver is male or female. The reputation information helps, but isn't always reliable.
>If someone is presenting themselves to you in person for entry into your bar, you have far more information to make a judgement on than the color of their skin... so it is not the same.
So the difference between "good" discrimination and "bad" discrimination is the amount of information on which the decision is based upon?
Logically then uber could add a "white only" option, "no queer" and "no leftist".
(of course this is arbitrary but you can easily come up with a reason why: if you split any group of real people in two it's only natural that one group has an higher incidence of a negative trait)
This also has a second problem: what if we let the passenger know not only the sex but also if the driver ate fish in the morning (and hundreds of other useless facts)? Does that make it discrimination because they have far more information?
I guess not but then how do you decide what information is valuable in order to decide if there is enough information to judge the individual instead of going off statistics?
How can you say that our theoretical racist patron is in fact racist and not going off the only valuable information?
No, that's a question.
I imagine it's not that since the rest of my comment is dedicated to pointing out how that'd be racist.
I was trying to make you explain what exactly the difference is since you didn't clearly define it in your reply.
Uber is also the one deciding to offer a rideshare service where mens are banned for working for them. Uber has the choice between vetting their employees and doing discrimination based on a correlated proxy. They choose the latter, and this discussion is about whether that is legal.
This same thing that keeps on happening when we try to reinvent things "without all that stuff that just adds friction." As with software, one should understand the underlying reasons for constraints in the old system before building the second one.
Banking -> crypto and NFT "without all that stuff..." -> wash trading.
Taxi service -> Uber "without all that employer stuff..." -> drivers with no background checks and no interview process
I understand part of this is routing around the damage of monopoly maintenance (medallion system, for example), but let's fix that instead of taking away the protections in place.
Sorry for the rant. I know this is like asking water to run uphill.
It happens with taxi drivers too. I know women friends/family that don't like going in taxis because of the unnecessary flirting and
harassment where with Uber it's easier to report and check by the driver's rating.
Hardly a rant. You're just describing the "move fast and break things" ethic (or should I say unethic). Or said another way: "all of the convenience with none of the responsibility."
So the only thing you're actually interested in is arguing "Trump bad."
Iran's religious leadership has been sponsoring terror throughout the world for the last 40 years. Jimmy Carter was duped by the British into causing the rise of Shia Islam there. Khamenei and his leadership had escape locations prepared in Venezuela. The U.S. rolled those up first, nabbing the leader of a criminal cartel (Maduro) in the process. Now Israel and the U.S. have taken out Iran's oppressive "supreme leader" at a moment in history when the Iranian people are struggling for their own freedom.
Khamenei was a bad guy. Maduro is a bad guy. They've put evil and harm out into the world and you're wringing your hands about it because it was Trump that stopped them?
"By G'Quan, I can't recall the last time I was in a fight like that. No moral ambiguity, no .. hopeless battle against ancient and overwhelming forces. They were the bad guys, as you say, we were the good guys. And they made a very satisfying thump when they hit the floor." -- G'Kar
But they've gone full-tilt Bozo with 11. The ability to deliver such an experience to their advertisers and marketing resellers was the whole reason for its existence. That is, 11 is about what Microsoft can get from its users, not about letting them use their own computer. It is no longer a suitable personal computing OS.
Warts and usability issues are present in Tahoe, and I wish Apple hadn't made the choices they've made on the UX, but Tahoe remains closer to being something for the user than Windows is.
Linux is generally for the sake of operating a computer (I didn't forget the user, but Linux--sweeps hand--makes the assumption that it's users are developers). That's what's so surprising about the age verification push in some Linux distributions (and their attendant bans for disagreement in the mailing lists or on GitHub).
reply