mtrovo

13 ms·

121.
by mtrovo 2y ago
Interesting read. Chasing perfection can be a real trap, as we are essentially trained to find flaws and improve things. That can easily turn into a never-ending cycle of fiddling with code without any clear end goal or time frame. One thin
122.
by mtrovo 2y ago
I think the author's point isn't that it's easier, it's the fact that being in an isolated, low judgmental environment allows you to express edgy ideas more uniquely without the prepackaged style you often find in big ci
123.
by mtrovo 2y ago
That's such a cheap take on the issue that it's hard to take serious. > The reality is that they would not be small – for example, the system being developed by Rolls-Royce is 470 megawatts, larger than most of the old, now clo
124.
by mtrovo 2y ago
I don't know if it's the rose tinted glasses of looking into the past as I'm getting old but I really feel we missed something along the way on the transition between desktop to web and now mobile. I remember how approachable
125.
by mtrovo 2y ago
Anybody with experience of using something like this with a big codebase and Gemini 2M context window? I tried a while ago (before 2.0 Flash) to solve some refactoring tasks and even after spending some time on prompt wrangling I didn'
126.
by mtrovo 2y ago
It looks like it's just a matter of time for "Computer Use" like tools becomes commoditised and widely available. I'm worried that this could upend our usual ways of filtering out bot activity with no simple way to go ba
127.
by mtrovo 2y ago
That’s the point of using AI in the first place. If your product is just a polished interface on top of a prompt, then your moat isn’t that strong, and chances are your product will be commoditized soon. By building a good UX and integratin
128.
by mtrovo 2y ago
We might live in a world where veto-ed assistants get VIP access to use the websites impersonating their owners without much second thought as long as you're at least on the paid Flash Max Pro™ plan.
129.
by mtrovo 2y ago
> Looking to 2035, BNEF’s global benchmark LCOEs falls 26% for onshore wind, 22% for offshore wind, 31% for fixed-axis PV and almost 50% for battery storage. Battery storage costs falling so dramatically is going to be huge. It's no
130.
by mtrovo 2y ago
> just $142/day I would love to see a deeper financial analysis of your scenario that consider the other variables mentioned like ops team, bandwidth, power, and cabinet rental. Also it's important to be honest about the initia
131.
by mtrovo 2y ago
I found the discussion around inference scaling with the 'Wait' hack so surreal. The fact such an ingeniously simple method can impact performance makes me wonder how many low-hanging fruit we're still missing. So weird to th
132.
by mtrovo 2y ago
First time hearing about the 90/50 rule and it's something I've seen play out so many times. It's easy to get caught up in just getting the code work on a limited case, but true quality comes from spending extra time on
133.
by mtrovo 2y ago
The fact that people are willing to pay more for AI features shows that consumer priorities are shifting and it's a good opportunity for Apple to increase its services revenue. That being said as a Pixel phone user I'm still not s
134.
by mtrovo 2y ago
I'm not American and I'm not politically involved on US politics, but I'm curious whether there's a solid rationale behind Trump's tariffs. My gut feeling tells me it might make sense if there’s a plan to onshore a
135.
by mtrovo 2y ago
The fact that both systems scored well on ARC AGI 1 shows they can handle unseen challenges without heavy human input, unless I'm missing something about why you see humans as the best interface for real world exploration.
136.
by mtrovo 2y ago
I'm now waiting for the clear definition of "sufficient expressive elements". It feels this is going to generate a heated debate about what qualifies as meaningful human input versus just hitting generate on a prompt. I think
137.
by mtrovo 2y ago
It's all about the right level of abstraction at the right time. The example given is too kind to the reader, I can imagine functions with 1000s of lines with the exact same structure of doing too much distinct sub-steps. Of course you
138.
by mtrovo 2y ago
The price comparisons there don't make much sense. DeepSeek V2 is far worse than 4o, so a better analogy would be to compare the pricing of V3 and R1 instead. For example, if you check https://artificialanalysis.ai/mode
139.
by mtrovo 2y ago
> There was also "a lack of transparent reporting with evidence that known weaknesses were not reported to key decision-making groups" and the authority was overly dependent on outside contractors. "The council did not equ
140.
by mtrovo 2y ago
The idea of treating AI agents as a distinct persona, similar to users or developers, is spot on. We really have to think about how they experience our software, this is essential for building effective systems that actually work for both h
141.
by mtrovo 2y ago
haha you got me. I'm real person using LLM to proofread the stuff I write. English is not my native language and I'm trying to improve my written vocabulary a little bit. Sorry if it reads a little bit too off.
142.
by mtrovo 2y ago
> it showcases Apple’s strength in AI That's a fanboy way to say Apple's strength is how long they can maintain their gatekeeping role. This moment feels like a pivotal opportunity for Apple to really ramp up its own AI offerin
143.
by mtrovo 2y ago
Wow, an 80% reduction in size for DeepSeek-R1 is just amazing! It's fantastic to see such large models becoming more accessible to those of us who don't have access to top-tier hardware. This kind of optimization opens up so many
144.
by mtrovo 2y ago
The most striking aspect of DeepSeek's success is how it challenges the prevailing narrative that cutting-edge AI requires massive financial backing. Their R1 model was trained with significantly less capital, yet it's apparently
145.
by mtrovo 2y ago
This is definitely making me rethink my choices. I've been eyeing the Model 3, but a 23% failure rate during inspections is really alarming. It's not just a minor issue if the car has three times more problems than other electric
146.
by mtrovo 2y ago
The use of LLMs to understand human-written content and extract structured data is something that has been long overdue. This will allow us to build more efficient and intelligent systems that can leverage the content already available on w
147.
by mtrovo 2y ago
More concerning is the practice of discarding the original audio recordings after transcription. This is a recipe for disaster. It removes the ability to verify the accuracy of the AI output and introduces serious ethical questions and it c
148.
by mtrovo 2y ago
Lumafield tool looks really cool, there are some links on a pinned comment to visualise the scans yourself. That being said I think comparing Apple’s Thunderbolt 4 cable to USB 2 cables and calling the others “imitations” is very biased and
149.
by mtrovo 2y ago
It's strange how we almost never praise people for earning their money. I remember growing up on a blue collar family on a blue collar city and being really proud of it. Instead now, we praise those who spend big, as if that is more im
150.
by mtrovo 2y ago
Norway's approach to electric vehicles really stands out. I've visited Oslo and it's so strange how the city feels so quiet with electric cars everywhere. Part of the reason is that they do not have a large ICE industry to pr

More ›