kyle_grove

Sounds like an experiment I'd like to give a try--what integration layer did you use to get LLMs to control the civ?

thecutline.ai , a Product Management Suite. I call it "A Product Manager That Says No", which stems from previous challenges I had using AI that was too sycophantic and optimistic to help with product decisions.

Working heavily right now on Customer Personas to use in validating/invalidating , which are configured with viewpoints, biases, and tendencies. Coming very soon will be Persona Journeys, in which you can get live, goal-oriented evaluation of your web app by a Persona.

Anthropic sign-on is surprisingly bad.

The problem of running a $4 Trillion consumer hardware company, with incredibly optimized supply chain operations, is that it heavily constrains the directions a new CEO would take the company, and by extension, the set of plausible people who could take the helm. I think even if the next CEO has a new or different product vision, they'd need deep knowledge on the hardware side of the house just to steer in any different direction.

I'd agree with all those facts about the competitive landscape, but in each of those competitors, there's enough wiggle room for me to think OpenAI isn't completely boxed in.

Google on multimodality: has been truly impressive over the last six months and has the deep advantages of Chrome, YouTube, and being the default web indexer, but it's entirely plausible they flub the landing on deep product integration.

Chinese companies and pricing: facts, and it's telling to me that OpenAI seems to have abandoned their rhetorical campaign from earlier this year teasing that "maybe we could charge $20000 a month" https://techcrunch.com/2025/03/05/openai-reportedly-plans-to....

Coding: Anthropic has been impressive but reliability and possible throttling of Claude has users (myself included) looking for alternatives.

Social: I think OpenAI has the biggest opportunity here, as OpenAI is closest to being a consumer oriented company of the model hyperscalers and they have a gigantic user base that they can take to whatever AI-based platform category replaces social. I'm somewhat skeptical that Meta at this point has their finger on the pulse of social users, and I think Superintelligence Labs isn't well designed to capitalize on Meta's advantages in segueing from social to whatever replaces social.

There's the technique of model orthogonalization which can often zero out certain tendencies (most often, refusal), as demonstrated by many models on HuggingFace. There may be an existing open weights model on HuggingFace that uses orthogonalization to zero out positivity (or optimism)--or you could roll your own.

Honestly, I think the mode that will actually occur is that incumbent businesses never successfully adopt AI, but are just outcompeted by their AI-native competitors.

IMO the lack of real version control and lack of reliable programmability have been significant impediments to impact and adoption. The control surfaces are more brittle than say, regex, which isn’t a good place to be.

I would quibble that there is a modicum of design in prompting; RLHF, DPO and ORPO are explicitly designing the models to be more promptable. But the methods don’t yet adequately scale to the variety of user inputs, especially in a customer-facing context.

My preference would be for the field to put more emphasis on control over LLMs, but it seems like the momentum is again on training LLM-based AGIs. Perhaps the Bitter Lesson has struck again.

I think in part because of YouTube demonization, which is how TikTok could poach the creators in the first place.

I suspect if they're mirroring content to YouTube, it's more to try to attract audience to TikTok than monetize through YouTube.

I would use the word 'fresh' for TikTok; like old school YouTube, there's quirkiness and variety.

My belief is that while eng manager empire building was the easier path to get promoted before 2022, it's not anymore, for two main reasons:

1. HC doesn't accrue like that anymore. 2. Many organizations are looking to delayer; harder to promote up to director when your org went from 9 runs to 5.

I hear a lot of the focus going to Tech Lead Manager roles--fewer reports but more hand-on keyboard than EM roles of the past.

As I understand it, the Q-hypothesis is often situated within the hypothesis of Marcan priority (Mark was the source for Luke and Matthew), and Q is a way of explaining agreements within Luke and Matthew that are not also found in Mark. The hypothesis would be that Luke and Matthew each combined text from Mark with Q.

I think (but cannot prove) that along the way, it was decided to explicitly measure ability to 'study to the test'. My theory goes that certain trendsetting companies decided that ability to 'grind at arbitrary technical thing' measures on-job adaptability. And then many other companies followed suit as a cargo cult thing.

If it were otherwise, and those trendsetting companies actually believed LeetCode tested programming ability, then why isn't LeetCode used in ongoing employee evaluation? Surely the skill of programming ability a) varies over an employee's tenure at a firm and b) is a strong predictor of employee impact over the near term. So I surmise that such companies don't believe this, and that therefore LeetCode serves some other purpose, in some semi-deliberate way.

I lived in Woodland for a time and I really wish I had heard about them back then so I could arrange a plant tour.

My 16GB M2 Air is doing it well.

Interesting, I'm playing with it and I asked it what SIEMs are and it gave examples of companies/solutions, including Splunk and RSA Security Analytics.

My main thoughts on RTO:

    1. No one work arrangement is optimal for all firms.

    2. Most policies around RTO (or remote work) are not meaningfully exploring optima (with respect to organizational health and work product quality).

    3. Therefore, the amount of preserveration and energy spent on RTO steals focus from main drivers for firm success. Which is probably the point, as the article points out.

I think your and nostrademon's comments are both insightful.

What I would add as someone who has been managing collaborative science teams embedded in large companies remotely, pre- and post-pandemic, is that some forms of alignment translate to the remote setting, but other forms of alignment are more challenged. I think the boundary is probably: if the teams were aligned pre-remote, you can sustain the alignment, even with new collaborative initiatives, but gaining new alignment with new teams is way more challenging.

Which is fine when what you are doing is Business as Usual, but falls apart when there are crises or disruptions that require net new collaborative relationships.

NordPass does the first part of this—it at least alerts you to breaches and prompts you to change it.

I think the full use case you describe is better suited for a passkey manager, since passkeys are marching generated already.

Mazda engine R&D is seemingly quite impressive: the SkyActive X engine is an unusual ICE gasoline engine that takes ideas from diesel, with greater fuel efficiency and horsepower.

I’m pretty confident (close to the 95% level) they will abandon the public charity structure, but throughout this saga, I have been baffled by the discourse’s willingness to handwave away OpenAI’s peculiar legal structure as irrelevant to these events.

The behavior of various actors in this saga indeed seems to indicate 'Altman and OpenAI employees back at OpenAI' as the preferred option by those actors over 'Altman and OpenAI employees join Microsoft in masse'.

BallmerGPT was developed before nucleus sampling.

It's possible, although contradicted by Brockman's statement, that Ilya voted merely to remove Brockman's board seat, and then was in the minority on the Altman vote.

I doubt this is what happened, but the reporting that Brockman was ousted from his board seat after Altman, and wasn't present in the board meeting that ousted Altman, doesn't make much sense either.

That will keep him employed for the three nanoseconds until he gets an offer from a competitor.

FOMO

Tell me, on which mention of “brother from another mother” did you stop rolling your eyes?

I don’t know that the telcos care one whit about degradation of trust in telephony; they’re effectively ISPs at this point and telephony is a legacy service.

In Athens, Ohio, there's Casa Nueva, a worker's cooperative Mexican cantina: https://casanueva.com/about-us/our-story/

In Ithaca, NY and its surrounds, Gimme! Coffee recently became a worker-owned enterprise: https://ithacavoice.org/2022/07/a-labor-of-love-gimme-coffee...

I’m generally in agreement with this. As this type of caution goes, some individuals could have enough experience and skill to know when they could relax on part of this, but for an inexperienced manager, not keeping distance is a giant potential pitfall.

One thing the post doesn’t exactly mention but is a very strong reason to be careful befriending your reports: favoritism. In my experience on data science and machine learning teams, it’s often the case that there are more and less glamorous work assignments on the team, and if you’re befriending of someone means they start to avoid all the undesirable work, you’ll end up damaging team chemistry and churning out good but unrewarded team members.