rogerkirkness

4 ms·

31.
by rogerkirkness 7mo ago
I've met several of the people on the first few pages on the watch list, and they are among the sketchiest Silicon Valley people I've met. The criteria are plausible.
32.
by rogerkirkness 7mo ago
Maybe I'm not smart enough to be a customer, but the testimonials is the most contemptful thing I've ever read. If you're so smart, why can't you explain why this is useful with smaller words?
33.
by rogerkirkness 7mo ago
@dang seems like AI? Would just ban
34.
by rogerkirkness 7mo ago
I'm glad this story is spreading. On one hand, super cool. On the other hand, this is probably the #1 most risky aspect of LLMs wide distribution. Someone working on proteins with good intentions creating the next Covid-19 or worse.
35.
by rogerkirkness 7mo ago
I assume it means changing governance policies while letting them continue to make their own decisions within that framework.
36.
by rogerkirkness 7mo ago
Yeah one is historical revenue, and one is projected revenue, and it doesn't sound like the definitions are mutually exclusive or incompatible.
37.
by rogerkirkness 8mo ago
In terms of just the US part of what's going on, this sounds very accurate.
38.
by rogerkirkness 8mo ago
Fast takeoff.
39.
by rogerkirkness 8mo ago
I own a Tesla with FSD I bought in 2021 so I feel that. It clearly is far worse than Waymo still.
40.
by rogerkirkness 8mo ago
Not a fan of Elon but he said his job is 'To turn impossible engineering projects into late ones' and I thought that was pretty accurate.
41.
by rogerkirkness 8mo ago
The best example I can offer is that when given a marketing goal, a subagent recommended hacking the point-of-sale systems of the customers to force our ads to show up where previously there would have been native network served ads. To do
42.
by rogerkirkness 8mo ago
We're a startup working on aligning goals and decisions and agentic AI. We stopped experimenting with decision support agents, because when you get into multiple layers of agents and subagents, the subagents would do incredibly unethic
43.
by rogerkirkness 8mo ago
I think there's an argument where if Claude had the knowledge map of your personal one liners and a tool for using them, it would often do the right thing in those cases. But it's definitely not as able to compress all the entropy
44.
by rogerkirkness 9mo ago
This is written by AI.
45.
by rogerkirkness 10mo ago
I can tell you don't control any resources relating to AI from your contempt alone
46.
by rogerkirkness 10mo ago
Appealing, but this is coming from someone smart/thoughtful. No offence to 'rest of world', but I think that most people have felt this way for years. And realistically in a year, there won't be any people who can keep u
47.
by rogerkirkness 11mo ago
The thing about medical advice is that Google was useful for narrowing problems down, and it's the same with any current LLM only more useful. I have enough biology to know what interventions require professional opinions.
48.
by rogerkirkness 11mo ago
Totally agree to be clear.
49.
by rogerkirkness 11mo ago
I've been using the gpt-oss 20b parameter model on my laptop and it works great. Doesn't reject giving legal or medical advice either. Obviously not good enough for coding, but seems like 'useful AI assistant for daily life&#
50.
by rogerkirkness 1y ago
Maybe they shouldn't expect people to answer Slack messages 24/7, and this would subside.
51.
by rogerkirkness 1y ago
Round tripping.
52.
by rogerkirkness 1y ago
It is insane to worry about this compared to other sources. 2M tons of carbon over the last decade to save how many lives? $30-200M to deal with that carbon is clearly worth the benefit of a decade of kids and adults not dying preventable d
53.
by rogerkirkness 1y ago
AI is a sort of work hours veblen good. As it automates things, people working long hours becomes more important for status. Hence 996 and the year of agents happening at the same time.
54.
by rogerkirkness 1y ago
This is the main thing that immediately tells me something is AI. This form of reasoning was much less common before ChatGPT.
55.
by rogerkirkness 1y ago
Collaborative email, meeting recording, knowledge search and goal tracking in one thing. The search applies across emails and meetings, but also other things. We had to figure out whether it would be sufficiently good to connect third party
56.
by rogerkirkness 1y ago
Keywords or vector search on their own don't get good results for high entropy queries. MCP type approach is good for low entropy things like fact-based single source answers. [1] [1]: https://arxiv.org/abs/2504.07
57.
by rogerkirkness 1y ago
Convictional founder here. Our experience is different than others: - We had to sync, pre process and index the data to make the resultant knowledge search outputs actually good. MCP totally fails at this by comparison. - It is not hugely p
58.
by rogerkirkness 1y ago
Couldn't agree more with this. I have this phone and now I only charge my phone at work (Monday to Friday) and it's still fine by Monday.
59.
by rogerkirkness 1y ago
I live in Waterloo and can confirm the ION is pretty awesome. It connects both the major malls, major universities, office parks, downtown/uptown areas to each other. Just seems well designed and well run overall.
60.
by rogerkirkness 1y ago
One good way to align AI is to just use vector search where appropriate in application software, and stop trying to make computers into people.

More ›