>No “vulnerabilities” in Tailscale were found or exploited, and that might make it even more uncomfortable for us. [...] But, we're a security tool. Their intrusion is our intrusion, and it's our job to take it seriously.
im a happy customer of tailscale, so i am obviously biased, but i have a lot of respect for this. they could have just stayed quiet and i dont think anyone would have bat an eye.
LLMs are writing as well as humans for people who think writing words is the goal. Printing presses also write words better than humans. LLMs are solving the problem one layer higher than printing presses, but still several layers down from the top.
Note that LLMs aren't useless, since they have other strengths which can make them equal or better than humans in certain applications. For instance, they can tirelessly review code for silly mistakes, making them useful in cybersecurity.
I've recently been using AI a lot for performance optimisation during a particularly busy period at work. I would say it was almost completely useless at the high-level direction - it would point out suspicious parts of SQL queries for example but on back to back testing these almost never resulted in any performance change.
In fact, if it wasn't for the fact that it made making the actual changes I identified much easier (move these joins into a CTE etc) it would have been a detriment. Not only did I get sidetracked by a bunch of useless suggestions but I also had to put up with others dumping their raw AI output at me as if it was somehow a meaningful contribution.
I really can't get over the way almost every piece of consumer technology/software has gotten worse in the last half decade. These companies are killing every goose they can get their hands on because they are so convinced there is some golden goose out there that will be worth more than all those regular geese combined.
The thing that makes it work really well is to make sure it has all the tooling to verify its hypotheses. If you allow it to run the full lifecycle in loops you will be surprised how well it works.
This is an important article. I hadn’t realized it was already getting this bad. Like a frog enjoying a nice warm bath ...
> Most people do not switch their operating system or phone provider every week either. But even if you do not utilize that freedom, it matters because it changes the relationship you have with the provider and the provider has with you.
This is why it’s important to utilize your freedoms. Do NOT let yourself get locked into a particular ecosystem (this is why I’m building a phone app for OpenCode).
This article makes me reconsider using my recently acquired Codex sub in my home setup. I never liked that they hide the reasoning, but somehow overrode the cognitive dissonance because the performance is so good. But the inauditability is already a huge problem.
Back in highschool, simulating different elevator algorithms was one of the projects I implemented during my CS class. It wasn't for the class — AP CS did not require anything like actual programming — but it was a fun project.
A cool connection is that a spinning-disk hard drive (HDD) is actually kind of like one really long elevator, just wrapped around a spindle instead of perfectly vertical. The SCAN algorithm is actually a disk-scheduling algorithm!
This is more exciting than k3, IMO. Dsv4 models are extremely cheap to serve. Improving their capabilities has lots of downstream effects, as it becomes "good enough" for more and more tasks.
DS was serving the pro version at extremely low prices for a long time, and they've had integrations with opencode & other providers, so they likely gathered a lot of data from real developers doing real tasks (on openrouter they were labeled as such). Now they can use those live scenarios to further post-train their models and improve them further.
Can't wait to see if distilling k3 into dsv4 brings additional improvements. Anyway, having fast cheap models getting better is great for the community. Especially since these don't "go away" on a provider's whim. Whatever capabilities they get, can be used "forever" going forward. And, at least flash can be ran "at home" with <10k in hardware, which isn't really possible / feasible with glm/k3 larger models.
I'm about to throw away multiple months of LLM generated code for one of my side projects. I was really careful writing design specs and it wasn't even a new code base the LLM worked on, but still after several months of AI changes I feel my code degraded more and more into a subtle mess. Hard to explain, each individual change looked good and logical and on the surface the codebase looks fine, but looking at the whole picture everything is subtly wrong in multiple ways. The same goes for where I used AI for existing commercial code bases. I would love to have AI write production ready software for me, but it's just not there yet, there simply are things that good programmers and architects do that cannot be captured by the training loop of current generation LLMs.
I notice the same pattern when using LLMs to write longer text like reports or scientific papers, individually each section they write makes sense but overall the whole document feels off in a hard to describe way. I think it's where you can see the difference between human intelligence and whatever it is LLMs have, it's not the same thing. We are much slower and less able on the small scale but seems we can do some higher level reasoning that is still impossible for LLMs. That always becomes clear when you point an LLM at an obvious flaw it produced and it goes "You are absolutely right!" as if it's obvious in hindsight but when running multiple "Please look for issues" iterations it would never have spotted the issue by itself.
That said I think it will be absolutely fine writing a simple CRUD app for you e.g. using some popular JS framework, Tailwind for styling and a regular ORM, there's more than enough training data available for these things. But then again such software could be purchased before already e.g. as a SaaS template, I don't think LLMs are so revolutionary here, they just replace the template (but to be honest a good hand-written SaaS boilerplate is probably still better than a vibe coded one).
Upsetting how we are reaching these lows while the administration is accusing everyone else of wasting taxpayer money except for themselves. At least under previous administrations you would get something for your money, like science funding and healthcare for the needy, not just bombing runs and posturing.
Wow, this article is super smart marketing by tailscale. Not only do they list all the nice and expensive features, that can help in such a situation but they also show that someone at huggingface made a very stupid thing by writing a reusable auth key in an env file. Everyone using mesh VPNs like tailscale, netbird etc. knows that this is like leaving the keys right at the door.
New Deepseek models are like Christmas for me. Really big fan of low cost API models, noone does it better than DS. Until VRAM price is low enough to run models locally, this is the way to go.
The subsidized subscription model won't last, API pricing "feels" closer to a true sustainable business model.
Great way to tank a brand. Part of BMW's value is/was the expectation of an upper middle class snob driving it, but having your car force ads on you is something people wouldn't even expect from a Kia Soul leased off a used lot.
I wonder if this is an artifact of how the author used random destinations. I worked in a building that used Destination Dispatch, and the common travel pattern seemed to be:
- Everyone who is not on the ground floor generally want to go to the ground floor.
- People who are on the ground floor generally travel in large groups to the same destination.
This happens because people who worked on the same floor often leave for lunch at the same time, and return at the same time to the same floor. Destination Dispatch helps in this case because it's batching large groups of people with the same destination.
This is not at the top as it is actively flagged by people that can't psychologically cope with the advances of AI. Hacker News is no longer a web site of an elite.
I use deepseek for a lot of my personal day-to-day agent needs, and I will simply put this here and let this speak for itself, last 30 days:
- Cost: $4.55USD
- API requests: 3,467
- Tokens: 323,183,886
And as an engineer who leads a small team, I have very high standards for quality, and these carry across to my personal projects where I use deepseek. It has not disappointed at all for coding or review tasks.
For everything else, use another model.
Not that I don't believe its possible to fix a lot of bugs, I also wonder what the actual dynamic was. Were the people in team working much more than usual as well? Given its Google, I wouldn't be surprised if there was an "internal push" to fix more bugs over next X sprints so that they can publish this blog and some manager can show impact and AI adaption to his superior.
To me this merely signals how broken C++ development really is. Most if not all of the bugs being uncovered are memory related and therefore intimately tied to the mental memory model of C and C++, namely manual memory management.
It's fine for a C or C++ program encompassing a couple hundred lines but beyond that it's a liability.
C and C++ are simply not fit for purpose when large scale software projects are concerned. All of these need to be ported to Rust or another memory-safe language ASAP to prevent mayhem.
The hundreds if not thousands of developers working on Chrome weren't idiots who didn't know what they're doing. The complexity of programming in C/C++ is simply beyond most intelligent individuals' ability to get perfect all the time.
I'll admit that I find this discussion a bit navel-gazy. It has become a question of semantics not a question of actual functionality. The question has become "what do we mean when we use the word 'reasoning'" which is uninteresting.
Dijkstra said[1] "... the question whether computers can think. The question is just as relevant and just as meaningful as the question whether submarines can swim."
I don't see a clear demarcation of the things that only "reasoning" can accomplish and can't be approximated or imitated by other methods, and so I think the question is simply not meaningful or relevant.
The Internet in the early 2000s felt a little more special as compared to today, when 99.999 % of content is locked in a handful of walled gardens. I mean sure everything we had then is still possible but let’s be honest everything about the web is fine tuned to deliver ads to our eyeballs, browsers, operating systems and even protocols are being designed with ad delivery in mind. Even the privacy friendly browsers just exist to serve ads when I think about it. Kind of sad really, but I guess you could be happy as it’s so big now and there’s so much to do. Just makes you wonder what stuff would thrive if ads didn’t exist. As of now every niche thing that gets successful will be invariably pulled into the ad ecosystem as it’s very hard to say no to money.
RSS wasn’t in the interest of the big platforms as it’s decentralized and there’s no good way to deliver ads through it, simple as that.
Not to denigrate the moment (AI ingress into theory which this is a part of) or the result here, but these headlines are perhaps overstating the importance - some of the theories and conjectures are available for AI-assisted exploration because they are quite niche and not very important.
Maxwell's name being invoked here for instance implies a hundred year old foundational problem like Fermat, but it's just a recent conjecture that was inspired by reflections from the great man on his work.
In a way the most remarkable thing about this is that it isn't even at the top of the HN homepage. Even if this is a step up from what we've seen before, we're no longer astonished by the idea that AI can make significant advances in mathematics and computer science.
What remains to be seen is whether Google also introduced more Chrome bugs in June than over the past two years, thanks to AI.
The big problem is that AI output can be very convincing and look "right", even appear to work, until you examine it in detail and realise all the edge-cases it didn't handle.
Quite the funny headline. It initially made me think that someone had come up with some sort of quantitative measure of the situational awareness of traders, and was claiming that there was an increase in traders making dumb trades that misread the situation or something.
Ironically, I would describe this selloff as an increase in situational awareness.
I've been driving flash model for 90% of my tasks. It's better than pro (for unknown reasons), very cheap and fast.
I try to keep changes under 1000 lines and drive architectural decisions myself, barely notice any difference compared to frontier models. The rest 10% is to spot bugs, security problems and to investigate better architecture, which flash can also do pretty well, I just cross check it.
Faster iterations are way better for me, I hate waiting for 5-10 minutes on small changes. I tried to use recent versions of Kimi and GLM, but they use too much thinking for no reason and are pretty slow because of it. I also often feed a lot of data to it, without worrying about hitting the limits: dependencies (to find bottlenecks in them), logs, performance dumps and so on.
Also, it will never complain about security guards, I've been using it to reverse engineer binaries.
im a happy customer of tailscale, so i am obviously biased, but i have a lot of respect for this. they could have just stayed quiet and i dont think anyone would have bat an eye.