sarabande

Seems like an unnecessarily broad brush to paint with, as if "tech" were a monolith. Just don't work for the companies that cross your self-stated lines in the sand:

> But my personal lines definitely extend to don’t give money to people who enable fascism and don’t create value for them, either.

I manage our company's model spend and build tools around it.

The good thing about the developer experience is that it is dead simple to create keys that are time-bound, model restricted, and have budgets against them in an easy, programmatic way (also in bulk). Costs are guaranteed to be on par or lower than any model provider, so it's an easy sell to the finance department.

To look at it from the other side, Gemini is the absolute worse when it comes to key management. The labyrinth of actions that one has to navigate to generate budgeted keys for a team is not worth it and I still haven't figured out how to manage the keys at a granularity I need.

OpenAI and Anthropic both have what I would consider to be middling DevEx. Most things work, but only on their platform, and there are some weirdnesses, like the fact that anthropic disabled its admin API key creation, so you can't programmatically create keys against a budget.

What happened then?

This is extremely difficult to read. Can someone summarize the point?

It would be helpful to show n. speakers of those languages or at least grade based on % of remaining non-English, non-Spanish speakers.

My intuition is that German, for example, is much less active of a language in the US, where immigration peaked long ago, than Chinese, where immigration is active.

Could you please consider making your points without "you'd have to be insane/an idiot to believe X"?

I found the article interesting, even given a large range of possible definitions of CTO.

I do wonder if it is possible to agree on a general definition of the CTO from the perspective of the job to be done, rather than how they do it.

For example, we could say the job of the CTO is to ensure the company remains technically competitive. If they do it by means of building an organization then so be it. If they rather do it by writing code themselves, then why not?

I really appreciated this blog post, John, to know that you're doing what I've been doing without a guilty conscience.

I'm a VP eng/research at a startup and also feel like one of the few people apart from the founders who can push major technical initiatives by just doing it themselves, due to: business context, technical chops, architectural judgment, grit, and seniority to pull in cross-functional stakeholders to help out.

However, I have often questioned if it is correct that so few people in the org can do this and if I shouldn't be enabling others to do it themselves instead.

How have you been able to navigate not having any direct reports? Who does your engineering org report to and how are you able to manage conflict between org builders and your technical vision?

Does anyone know if they published the dataset?

I'm curious about the simple, lightweight change management process listed in the article. Anyone have any tips on doing this well? I kind of hate the part of having to convince everyone of something "obviously better" :)

I enjoyed this piece, including references to the Stoics and Spinoza. It preaches serenity, goodwill, composure, etc.

As someone in their 30s with children, work and a generally busy life, I wonder if anyone can recommend some pieces with more direct application - that is, in this vein, but perhaps an operational / how-to guide. Sometimes, it's hard to translate principles to action.

Consider outbound AI calls as a weapon against inefficiency. What is the game-theoretic equilibrium when faced with adversaries like call centers & other businesses with labrynthian customer service?

I bet it leads to more efficiency for everyone. When inbound robocalls deluge a business, the business pushes to cost-optimize its own service.

Business replaces humans with similar AI solutions to handle the phone modality, but hopefully then reverts to great service via email/API to reduce costs further.

Then, humans using AI voice services can "de-escalate" and revert to email/API AI, e.g. going from:

1. Business: AI (Voice) or Human | Customer: Human

2. Business: AI (Voice) or Human | Customer: AI (Voice)

3. Business: AI (voice) | Customer: AI (Voice)

4. Business: AI (Text) | Customer: AI (Text)

Can you add calendar week and fiscal year start

I used the example sentence from your link.

纔 in this case should use the definition of 才 (cai2) not (shan1) which is extremely uncommon. Otherwise, cool app!

Can you share more about your experience regarding how management at a small company lead to management at a big company? Was the switch voluntary, and how did you do it?

"Absolutely ruined"? We still get value from flight, just now trade wait time and privacy for security. It's clearly still worth it for the vast majority of the world, see global flight stats.

Multiple EU airports like Schiphol already do facial scanning+passport for automated entry for citizens of select countries; so does CBP on US entry from abroad. What's the issue, that it's a subcontractor?

[This is not legal advice]

Not a good sign. Most DE-based eng contracts max at 3 months, or the default per § 622 BGB.

6 months likely means the company fails to retain good staff. Negotiate for asymmetric § 622 BGB notice for yourself, and 3 months notice from the company.

I like the article's premise. I would like some argument for why we'd suddenly get better at measuring something that has escaped us for at least 80 years when knowledge work became more common.

> My suspicion is that [...] there will be a surge in our capacity to measure and evaluate real-life work performance.

It's your personal preference to be addressed with _Sie_, but languages change and German is seeing shift to the use of _Du_ even for transactional business [1]. Target audience, region, familiarity etc. will all affect which word is used. Claiming the use of _Du_ in customer relationships as "wholly inappropriate" again is an expression of personal preference, not fact.

[1] https://www.abendblatt.de/wirtschaft/article233949947/firmen...

I also didn't know, some Googling led to this being the likely reference: https://en.m.wikipedia.org/wiki/OneWeb#2022_Russia_controver...

Here is a summary, part ChatGPT and part me:

The “Lone Banana Problem” describes subtle biases of Large Language Models (LLMs) in AI: LLMs reproduce the statistical average of the inputs that they have consumed in the context of the question they have been asked. It's called that problem because the model used to generate images has never seen an individual banana, so when prompted always generates two bananas.

FYI: A quine is a computer program which takes no input and produces a copy of its own source code as its only output.

I'm trying to use this on a 3M mp3 file to test ASR with language code deu, CPU only, and I keep getting this error -- are there limits to the MMS inference?

  File "fairseq/data/data_utils_fast.pyx", line 30, in fairseq.data.data_utils_fast.batch_by_size_vec
    assert max_tokens <= 0 or np.max(num_tokens_vec) <= max_tokens, (
  AssertionError: Sentences lengths should not exceed max_tokens=4000000
  Traceback (most recent call last):
  File "/home/xxx/fairseq/examples/mms/asr/infer/mms_infer.py", line 52, in <module>
    process(args)
  File "/home/xxx/fairseq/examples/mms/asr/infer/mms_infer.py", line 44, in process

I don't know about in the US, but I suspect it is similar to Germany, where the bar exam also requires you to argue/reason about/solve novel cases, i.e. not regurgitation of stored information. The AI likely wouldn't pass that.