827a

8 ms·

61.
by 827a 4mo ago
Companies have never cared about security, because there are almost no consequences to data breaches. A hospital network could get ransomwared for 48 hours, and no one cares. Critical data gets leaked? So what, pay a fine. You either pay a
62.
by 827a 4mo ago
I would postulate that there are two reasons why this is happening. 1. Pessimistic, harsh, etc: the quality of US graduates has been falling. Reading comprehension has been on a downward trend over the past decade. Mental illness, depressio
63.
by 827a 4mo ago
How do you know your skills aren't marketable if you don't go out and market them? Go apply around. You don't necessarily need to leave to do it.
64.
by 827a 4mo ago
Didn’t the UK not six years ago vote to leave the EU? Is he suggesting that there’s demand from Britain to reverse that? Or is he suggesting they be forced to rejoin?
65.
by 827a 4mo ago
This week an S&P 20 company with previously unlimited Claude limits also set a $250/mo/person limit; though its unclear to me how widely the limits are being enforced, may be the case that its just non-software engineers. Do w
66.
by 827a 4mo ago
Cringe
67.
by 827a 4mo ago
I believe the correct way to interpret AISI’s findings is that both Mythos and 5.5-Cyber are capable of solving their full benchmark (the only two models that can); Mythos does it with fewer tokens and more consistently. Two things of note:
68.
by 827a 4mo ago
GPT-5.5-Cyber has already at least hit if not surpassed Mythos capability in cyber tasks. The only reason they're holding back is because once its out everyone would realize that its capabilities were a step change in March, but are no
69.
by 827a 4mo ago
Though in fairness: Windows on ARM (even ignoring the earlier iteration of it from ~2016) has had a year or two to mature thanks to the Copilot+ PCs. And, this particular chip seems near-identical to the one Nvidia put in the DGX Spark (aga
70.
by 827a 4mo ago
Repayment isn't a goal that anyone in the system should reasonably want. Federal debt is not like credit card debt. Debt is a product that the US Government sells. Me, being a big corporation or human, go to the USG and say "I n
71.
by 827a 4mo ago
The $175T number is unfair because it treats Social Security and Medicare/aid as a liability instead of the service that they are. You might as well say the US is in infinite debt, because we'll always be paying something for our
72.
by 827a 4mo ago
jq can 100% be an MCP tool. Remember: Agent tools do not have to involve a network boundary. They can be natively implemented inside the agent harness, and/or they can be provided via a local MCP server. The point of making it a tool i
73.
by 827a 4mo ago
Cursor is functionally "one of the big 4 labs" (SpaceX).
74.
by 827a 4mo ago
Cursor only supports a single model (Kimi K2.5) not made by the Big 4 labs (OpenAI, Anthropic, Google, xAI). Cursor is actually extremely bad at wide model support. OpenCode is much better at it.
75.
by 827a 4mo ago
You’re right that having a shell is the ultimate tool, and an agent with a shell seems to perform better than one without one. But, making shells safe is really damn hard; e.g. in the context of running an agent on behalf of a SaaS customer
76.
by 827a 4mo ago
Yes, in the same way a programming language would be worse off if they focused all of their effort on building an implementation instead of a language specification. You could literally, deterministically, zero AI, code-gen a CLI from an MC
77.
by 827a 4mo ago
The idea that MCP tool definitions take up a certain number of tokens is laughable. That's an implementation detail of the agent harness. MCP is just an API specification. Hell, there's nothing in it that makes it much of any diff
78.
by 827a 4mo ago
I definitely believe that you can discern differences between Opus 4.6, 4.7, and 4.8. I might also believe that you believe that you can discern improvements between Opus 4.6, 4.7, and 4.8. But conclusively, consistently, scientifically,
79.
by 827a 4mo ago
You're not successful until you exit. And, of course, there's always room to be more successful.
80.
by 827a 4mo ago
I never said they were doomed. Where did you get that idea? I said they aren't ready for this world. That means they screwed up and need to get ready. They let the Mythos hype get to their heads while the world changed beneath them.
81.
by 827a 4mo ago
We'll have to agree to disagree on that last point. I think that, historically (past ~6 months), "always use the most advanced model" being the norm is really just an artifact of both: The most advanced models oftentimes bein
82.
by 827a 4mo ago
Agreed. But we're seeing Cursor (now SpaceX) take these models and add great coding capability on top of them. Frontier model providers should be concerned that Composer 2.5 costs $0.50/$2.50 (versus Opus 4.8 $5/$25). That&#x
83.
by 827a 4mo ago
Frontier models are mostly past the point of human ability to discern whether they are actually better or worse than predecessors and competitors. I suspect the benchmarks may also be saturated, or at least past their usefulness. I personal
84.
by 827a 4mo ago
Companies that are brave enough to do this; or more importantly, those that have aligned their internal processes and infrastructure around an ability to do this; will own the future. The scary part for some existing companies is that the s
85.
by 827a 4mo ago
Anderson Cooper: But predictive markets do rely on someone having some inside information. Shayne Coplan: Uh-huh. Yeah. I think that people going and having an edge to the market is a good thing. Obviously, you need to curate them and you n
86.
by 827a 4mo ago
I'm well aware (no one is using it there, either).
87.
by 827a 5mo ago
[flagged]
88.
by 827a 5mo ago
Yup; its not just that people want the best models, so they use Opus or GPT-5.5. Its also that we're talking about nationalized chinese companies. Western corporations are not en masse forming business relationships with Chinese firms
89.
by 827a 5mo ago
The issue you're not seeing is: Western corporations, the primary drivers of AI spend globally, are not forming business relationships with nationalized Chinese AI labs in order to directly use the DeepSeek API. They're using it t
90.
by 827a 5mo ago
In the statement "we got half a percent of Google's marketshare with a model we're literally giving away for free" the term "we're" here refers to the conglomeration of "DeepSeek" (for making a m

More ›