bob1029

33 ms·

511.
by bob1029 5mo ago
I think web frameworks have been "in trouble" as of gpt-5.4. I can't imagine using something like React anymore. The most incredible combo I've seen lately is progressive enhancement of Razor Pages with javascript. With
512.
by bob1029 5mo ago
> Our findings reveal a phenomenon of constraint decay: as structural requirements accumulate, agent performance exhibits a substantial decline. I have exactly the inverse findings on my end. The bigger and more legacy the codebase, the
513.
by bob1029 5mo ago
Seed oils probably aren't ideal, but I also worry how much this narrative is distracting from the bigger problem of sugar consumption. Humans only have so much attention and discipline. It would be a shame to focus all that energy on a
514.
by bob1029 5mo ago
I do think token demand might actually become satiated at some point. Machine to machine scenarios imply the potential for infinite sprawl, but machines don't run for free or forever. There is always a human with interest involved some
515.
by bob1029 5mo ago
C# gets all the mindshare because it's easier to understand and use on average. We can all agree that F# is more clever and concise. No one is dying on that hill. But in terms of hacking your way through the customer requirements and w
516.
by bob1029 5mo ago
Winforms is better than ever. You can use it with .NET10 and WebView2 is a thing now.
517.
by bob1029 5mo ago
I think JWT-like schemes can enable better machine-to-machine integration. If you drink all of the koolaid, you can wind up with a system where two different parties securely authenticate sessions without any kind of secrets ever needing to
518.
by bob1029 5mo ago
I think quantifying tokens used is analogous to quantifying the amount of sawdust generated on a construction site. Churning out useful code quickly is not solved by using more tokens per unit time. Most non-technical leaders can grasp th
519.
by bob1029 5mo ago
I believe speed was the biggest factor. You can go a lot faster when you aren't contending with traffic. Vertical = multiple horizontal planes that don't interact much.
520.
by bob1029 5mo ago
Any kind of fixed capacity usage model seems to be a dead end. Paying per token might seem like an exploitative arrangement at first glance, but it's a luxury if you are experimenting or deploying greenfield. Provisioned capacity is a
521.
by bob1029 5mo ago
I am a bit suspicious of these ideas. When I disabled parallel tool calls in my custom gpt5.4 harness, the quality of results went up dramatically. It looks like it's running slower and it probably is for some problems, but it's c
522.
by bob1029 5mo ago
I worked at Samsung when they just got done transitioning from rail-guided vehicles to an overhead handling system. The automation rate easily compensates for any specific end to end delays. You can have thousands of test wafers in flight
523.
by bob1029 5mo ago
I think the Death Star is the most apt analogy so far. You can either help build and maintain it, or you can risk becoming one of its first test targets. In this analogy, the laser system has demonstrated to function at low power as of a fe
524.
by bob1029 5mo ago
This kind of indirect exposure might look good on paper but it's never even remotely a linear mapping in practice. Holding the underlying directly is the best bet if you want to minimize the possibility of getting screwed over by exter
525.
by bob1029 5mo ago
The biggest issue with the interim onsite generation is the lack of meaningful stack height on the generating units. Airplanes by virtue of their mode of operation stay out of the unhappy regime most of the time. Also, engines at/near
526.
by bob1029 5mo ago
The non-linear scaling on the slider is an excellent UX.
527.
by bob1029 5mo ago
The Houston metroplex might be one of the best domestic examples of the urban heat island effect. They've got their own entire website about it. If you overlay the daily temperature curve of 77002 with any zip outside the beltway, the
528.
by bob1029 5mo ago
I think it's more about the popularity than the capability. The chances you might accidentally put a Github access token into an undesired security context goes up dramatically when you actually create and use one on a regular basis. T
529.
by bob1029 5mo ago
It definitely seems like the point of no return has been passed. The size of the codebase doesn't matter anymore. In fact, I am finding that the larger the codebase the better the performance. Starting from scratch with vague ambition
530.
by bob1029 6mo ago
I think it will. The primary issue is simply that developers are the most immediately impacted by this technology. The combination of being able to adopt, willing to adopt, and the tech actually being incredibly good at developer related co
531.
by bob1029 6mo ago
The longer I look at the AI transformation, the more it seems like a people problem than a technology problem. The technology is undeniably there. The people are all over the place. I am watching a 10 person company try to run 3 different A
532.
by bob1029 6mo ago
Anything that can't be explained and iterated deterministically is too risky for the business of declining financial transactions. Human analysts need to be able to explain to compliance in a single 5 minute email why a specific transa
533.
by bob1029 6mo ago
> billions of lines of code. Billions as in 10^9?
534.
by bob1029 6mo ago
A combination of SQL and git seems to work best. They tend to complement each other really well. git.exe can't tell you things like how many references a specific type has. It can get close with grep and friends, but it's not very
535.
by bob1029 6mo ago
The relative and auto-scaling nature of Elo ranking feels like an advantage here. Relative ranking systems extract more information per tournament. You will get something approximating the actual latent skill level with enough of them.
536.
by bob1029 6mo ago
In terms of "direct" retrieval strategies, the most powerful approach I've seen so far involves using git.exe. The available commands like grep/log/show/blame/diff/reflog/bisect cover a lot of gr
537.
by bob1029 6mo ago
Audio amplifiers can be strongly affected by noisy waveforms. Class D amplifiers and other topologies that depend upon SMPS for power delivery are usually unaffected. Class A/B is where you will typically hear it.
538.
by bob1029 6mo ago
The best strategy is to frame your argument from the perspective of the customer: > This will allow for us to deploy the feature in only X days supporting Y use case with Client W who has been complaining about this shit for Q months now
539.
by bob1029 6mo ago
Python might still be the best option if your goal is to perfectly one shot the solution and minimize token usage as much as possible. However, if you are willing to stub your toes, retry, and pay more money, an entire new world opens up. L
540.
by bob1029 6mo ago
> Is that via a MCP server? No, this all in one application. A Winforms+WebView2 app wraps the chat completion APIs and implements the various tools directly.

More ›