lumost

4 ms·

31.
by lumost 2mo ago
how do you optimally balance code generation and virtual function calling patterns for the best performance? Is there a standardized OLTP format on object storage? How do you performantly implement differential privacy policies within the d
32.
by lumost 2mo ago
I'd ascribe "deep tech" to anything that you can reasonably get a PhD in and have it not be unusual. There are dozens of academic conferences on DBs pushing the frontier forward.
33.
by lumost 2mo ago
Deep tech fields have a more complex relationship with academia than other industries. Engineers often need license to test ideas that may not see adoption for years, or require tens of millions to build-out. A "research" arm is e
34.
by lumost 2mo ago
If you are asking a human to review something, it should have been verified/reviewed by a human first. I have no interest reading someone else’s ai output that has not been verified.
35.
by lumost 3mo ago
There will be cloud/SaaS vendors who have lower cost of labor/capital due to automation and financing terms. Having these models in the open caps the inference margin.
36.
by lumost 3mo ago
We also may be at the wealth inequality level where prices become… weird. If there is really only a few dozen people doing the buying and the selling at the top on a weighted basis, then the prices are whatever they convince themselves of.
37.
by lumost 3mo ago
These companies have valuations reflecting a debt light business. At a minimum, 420 billion in debt is enough to change the stock price by 10-20%. If the company plans to add another 400 billion in debt you need to give it the side eye. If
38.
by lumost 3mo ago
yes, above 50% I don't observe a significant improvement, the models scoring above 65% like opus4.8 seem worse than gpt5.5/opus4.6. DeepSWE appears to align more closely to my personal experience.
39.
by lumost 3mo ago
Does this even look like a timing race at this point? Moonshot only really got going this year. Why couldn’t any well capitalized firm get to a leading position?
40.
by lumost 3mo ago
Honestly my anecdotal experience is that fable is benchmaxxed. I have not observed useful gains for Claude since 4.6, with each model iteration making progressively poorer decisions in pursuit of its goal. The 5.5/5.6 series has perfor
41.
by lumost 3mo ago
The nature of ML is to reach for easily measured proxy problems which can then be optimized. Why worry about a backoffice integrated micro-service you have no data on when you aren't sure that you can design and implement a rewrite of
42.
by lumost 3mo ago
There is a big divide between “application ai” and “model ai” startups. Model ai startups start from OSS models, and use them extensively for different purposes as their work would usually be banned by proprietary labs. Application ai start
43.
by lumost 3mo ago
The fear is that we can’t tell when the ai advice is bad on these subjects, and as such probably accept confidently terrible advice. How often do managers just regurgitate ai advice rather than consulting their experts? How often does a per
44.
by lumost 3mo ago
I unfortunately have to use grok via tesla. The grok voice chat is objectively decent.
45.
by lumost 3mo ago
I think most nuclear folks would rather divert the "big coal, and nat gas" plant building budgets to "build nuclear." I understand the motivations for solar/wind, but there are real limiters that aren't address
46.
by lumost 3mo ago
This is a potentially strong indicator of how institutional investors view spacex. Given that Spacex is in a high depreciation/capital intensive business, a high cost of capital and potentially difficult capital terms is problematic. S
47.
by lumost 3mo ago
The risk here is that either AI commoditizes the software "why ask duolingo to ask chatgpt when I can do so directly?" or simply adds cost to an existing service "why does every costco employee now cost 10% more?" So far
48.
by lumost 3mo ago
I do wonder if the productivity gains control for increased bug findings and token costs.
49.
by lumost 3mo ago
I think the counter point for these projects is that you may not need a deep understanding if you can measure the outcome. While this may not be true every time today, it plausibly will be in the future - making the activity worthwhile.
50.
by lumost 3mo ago
When I was in a large org, it was a common failure that a manager would come in from a smaller firm - and assume that our design practices etc. were meaningless overhead. Inevitably they'd try to do a "rapid prototype" of a f
51.
by lumost 3mo ago
I think the theory is that they can expand the infrastructure enough that conventional fiber etc. stops being competitive.
52.
by lumost 3mo ago
if super chargers were always within 5 minutes, located around desirable locations to spend an hour, did not charge idle fees, and everyone had 30 minutes of free stuff to do at these locations... then it might just work!
53.
by lumost 3mo ago
This math is bad and shouldn't be used as a justification. It presumes that every charger is always available, that variable battery conditions and traffic conditions make reaching that charger every time, and that drivers have time to
54.
by lumost 3mo ago
the counter point is that building or selecting the specialized model may cost as much as the lifetime inference costs of the task with the specialized model. If I need to pay someone 300k to make the model and infrastructure... then I woul
55.
by lumost 3mo ago
remote inference should be sufficient for most robotics applications with potentially a small model for safety critical actions running locally. Unless you are in military robotics or automotive of course :)
56.
by lumost 3mo ago
The Niche model story is still fairly week. Evidence points to general models being equally capable to niche models at a more attractive capex (risk is spread across multiple verticals rather than concentrated in a single model capability)
57.
by lumost 3mo ago
I suspect the concern is that model serving is a stateless “simple” problem. As yet, no one has identified a reliable moat in inference. If the moat is performance, then prices will collapse. Unlike traditional cloud moats around state, ope
58.
by lumost 3mo ago
If I understand correctly, there are more people employed in manufacturing today than there were a century ago. The hollowing of western manufacturing due to policy choices created a false perception that the sector was in decline globally.
59.
by lumost 3mo ago
it's a common missconception that engineers spend most of their time producing code based on documented requirements in jira tickets. I'd believe that a complete automation of this aspect of our industry would only be enough to pr
60.
by lumost 3mo ago
so all we need is someone to leak a sufficiently large amount of claude generations onto the open and private web for all other LLMs to mimic the same marking style? wouldn't this happen due to the massive amounts of spam/slop bei

More ›