Hacker News: amsilprotag

New comment by amsilprotag in "US Job Market Visualizer"

amsilprotag — Tue, 17 Mar 2026 01:04:32 +0000

I thought the same as gp, that putting teachers at high risk invalidates the whole visualization. If this is intended to be useful for future career planning, with meaningful gradations between specializations, than it should exist in the probability space where human agency still matters. And in that space, from a Riccardian and political economy perspective, high human-touch jobs with strong public unions should be among the safest.

New comment by amsilprotag in "How London became the rest of the world’s startup capital"

amsilprotag — Thu, 29 Jan 2026 16:05:29 +0000

Any time you use the word startup you are already ceding ground. Startup as a term solves the problem of the VC: given technology is inherently risky, how do I get good returns? Answer: make sure the winner keeps doubling down. Create a culture around valorizing the few big, big winners and writing off-- financially and mentally-- the losers.

14 years ago now: https://www.paulgraham.com/growth.html

Of course, this culture is less helpful for individuals and cities, who have to live with the volatility.

New comment by amsilprotag in "Learn Your Way: Reimagining Textbooks with Generative AI"

amsilprotag — Thu, 18 Sep 2025 21:48:08 +0000

I have used the quiz-making learning tool within gemini. It is very good for things that would exist in a typical K-12 textbook. The first 30 or 40 multiple choice questions on a subject are usually pretty good and useful. But then it will tend to repeat multiple choice answers, give strictly wrong answers, repeat questions, or offer multiple valid answers. The answer explanations are what you'd expect with little human QA. Still a useful tool for people who sanity-check the given answers, but it might do more harm than good if people don't follow up on their confusion.

New comment by amsilprotag in "Duolingo CEO tries to walk back AI-first comments, fails"

amsilprotag — Mon, 26 May 2025 19:25:26 +0000

Maybe you are right that I give too much benefit of the doubt. I have been following the saga and I believe the AI turn was a bad move on every level. That's why I stopped using the app. But I can still believe that the CEO doesn't want to ship the cluttered garbage that is the present app. I think the pop-ups are a net-negative even purely financially with churn outweighing subscriptions. So my model for the situation is that he spent his credibility on AI, which was bad, and now doesn't have the credibility to spend to change the metrics that guide every team's behavior, so the company decays entropically. Maybe the clutter comes from the CEO trying to up subscriptions, but based on the first 10 years of the app, I believe he has better taste than to do that, so I look for a more complex explanation. Again, may be giving too much benefit of the doubt.

[edit: thinking about it more, I think I have built up a lot of goodwill with the app over the years, and it's a strange mental process for years of goodwill to evaporate over the course of a few weeks]

New comment by amsilprotag in "Duolingo CEO tries to walk back AI-first comments, fails"

amsilprotag — Mon, 26 May 2025 18:58:32 +0000

A CEO has the power to do anything, but employees have the power to collectively, quietly sandbag if they don't like the leadership. I think the AI effort led to a broad disillusionment, causing an unwillingness to put extra effort into their work. Across the company, everyone starts to take the path of least resistance. The CEO senses his influence waning and becomes more accommodating to avoid further morale death spiral. So situations can arise where a CEO would like a cleaner product (no one likes to ship garbage), but has lost the political capital to make it happen.

New comment by amsilprotag in "Duolingo CEO tries to walk back AI-first comments, fails"

amsilprotag — Mon, 26 May 2025 18:34:36 +0000

I stopped using duolingo regularly about a month ago. It's wonderful that Luis von Ahn says in interviews that he tries to prevent teams from cluttering the app, but it seems like he lost the battle. You can get 10+ pop-ups after a lesson. The friend feed is cluttered with meaningless achievements. The web app is tolerable, but the phone experience is miserable. But if you're behind a computer and keyboard, there are much more effective ways to learn. Busuu is a much warmer product on either device, with videos of native-language speakers to help with listening.

Duolingo has scaling and distribution. It makes no sense to scrimp for pennies on a product (e.g. English learning Spanish) that has millions of daily users. The AI radio lessons feel alienating and demoralizing compared to voice-acted stories, and the quality control is much worse.

New comment by amsilprotag in "Liu Cixin's Technologies of the Future"

amsilprotag — Fri, 13 Jan 2023 16:05:38 +0000

Totally agree with poor plot and dialogue. But if you had read it on publication in 2008, viewing the aliens as US and Earth as China, you would have been one in a small minority to foresee something like the Oct 2022 Biden order to revoke citizenship from anyone working in the Chinese semiconductor industry.

New comment by amsilprotag in "Twitter Deal Temporarily on Hold"

amsilprotag — Fri, 13 May 2022 11:29:00 +0000

I don't know how well this would hold up in arbitration, but he has repeatedly cited the existence of bots as a reason for buying and fixing twitter.

> If our twitter bid succeeds, we will defeat the spam bots or die trying!

April 21 900K likes

https://twitter.com/elonmusk/status/1517215066550116354

New comment by amsilprotag in "Early Work"

amsilprotag — Tue, 20 Oct 2020 14:22:48 +0000

I don't know whether this is true, and I'm writing this out of curiosity and not bad faith, but the most parsimonious explanation for this being the top comment is that both the writer and voters assumed (as I did before reading) that the essay was about time-of-day work rather than time-of-project-lifespan work.

But it's a good point! Faith fluctuates by day, and those low-faith days are when a small project is abandoned. I think graham's solutions (supportive friends, ambitious city, historical examples) are a good way to hold the faith when the general public and the project itself don't seem to warrant it.

New comment by amsilprotag in "If founders treated their investors the same way they treated their employees"

amsilprotag — Tue, 18 Aug 2020 21:46:06 +0000

Preferred stock? Priority?

New comment by amsilprotag in "Epic direct payment on mobile"

amsilprotag — Thu, 13 Aug 2020 16:59:20 +0000

Tim Sweeney is really interesting. He coded Unreal Engine 1 and is controlling shareholder of Epic. He gave a speech[0] January at a game conference arguing against the App/Play store 30% mark. I think he has internalized Ben Thompson's FAANG strategy concepts and created the best denunciation of the mindset, in this speech attacking the concept of "owning the customer." He joins Shopify in trying to create platforms in the Gates sense where most of the money goes to consumer and 3rd party. And a lot of his actions, in the short term, are against Epic's financial interests, commodifying and cannibalizing their own dominant market position for the sake of growing game industry TAM and a vision of future interoperability[1].

[0] https://www.youtube.com/watch?v=CJFO2k1-tyg

[1] https://www.matthewball.vc/all/epicprimer1

New comment by amsilprotag in "Day Two to One Day"

amsilprotag — Tue, 17 Sep 2019 18:23:36 +0000

I recently finished Brad Stone's The Everything Store, written in 2013. In the last chapter, brands complained about Amazon pricing below the their Minimum Advertised Price (MAP), to the detriment of physical sellers. Stone used the German knife company Wüsthof as an example. Wüsthof at one point around 2010 stopped supplying Amazon, but its products were still sold by third parties. Looking at Amazon now, there are at least 100 in-house Wüsthof products being sold by Amazon at prices that seem lower than the mid-hundred dollar price ranges cited in The Everything Store.

Does anyone know what happened generally to Amazon's relationship with brands over the past six years? Presumably over this time, Amazon has only gained in bargaining power over brands at the expense of physical locations, in accordance with Ben Thompson's Aggregation Theory. As an outsider, it seemed like Amazon and brand buyers have done well at the expense of brands and retail stores.

New comment by amsilprotag in "M.I.T. Media Lab concealed its relationship with Jeffrey Epstein"

amsilprotag — Sat, 07 Sep 2019 17:50:55 +0000

I think Giridharadas is very good and says good things and picks good targets. But there is a strange attitude here and on twitter that he is guileless, when clearly he is making very calculated moves. This is especially apparent in seeing the Hoffman comment as a threat but not the resignation letter. Maybe the gp came off as harsh, but I'm trying to explain why the ggp should not expect the husher line of attack to be immediately adopted.

New comment by amsilprotag in "M.I.T. Media Lab concealed its relationship with Jeffrey Epstein"

amsilprotag — Sat, 07 Sep 2019 17:35:24 +0000

From your text: Then Anand shares the thoughtful and profound email he sent to Ito & Co and others at MIT Media Lab, from which it is crystal clear Anand is unambiguously trying to do the right thing.

You're giving Giridharadas way too much credit. His e-mail resignation is an implicit ultimatum (if you don't fire him, I will go public), to which Hoffman counters by saying how he will try to frame Giridharadas as trying to personally gain from this scandal (which, clearly he will).

By publicly stating Hoffman's weak threat, while reserving the future threat of revealing their correspondence, Giridharadas makes it harder for Hoffman to follow through on the threat without public backlash.

So Giridharadas targets this tweetstorm to inflict maximum damage to Ito supporters and MIT admins. The hush money line is harder to prove and channels frustration towards the hushers more than the facilitators. The cash for prestige line places all the blame on Epstein and Ito. Then he publicly shares the list of Ito supporters [0] and creates public knowledge of their thinning numbers. When the firing eventually occurs, Giridharadas will have collected the spoils of a successful cancellation, the public knowledge that he can make very credible threats.

If you are correct, we should expect Giridharadas to place more emphasis on the hushers for the sake of justice. If I am, we should expect him to put all effort into bringing down Ito.

[0] https://twitter.com/AnandWrites/status/1170025167063605250

New comment by amsilprotag in "Startups from Y Combinator’s S19 Demo Day 2"

amsilprotag — Wed, 21 Aug 2019 18:56:01 +0000

This sounds like what Kevin Kwok was describing[0] in his recent essay, The Arc of Collaboration. Even down to using the Discord metaphor. Kwok says that Slack ended up being more of an exception handler when normal processes break down. Contrast to Discord, which is more of a meta-layer around games and integrates within game platforms. I wonder if Tandem is what will make collaborative communication ubiquitous, i.e. sales talks to engineering talks to design all from within their various environments.

From TC: Tandem provides a virtual office for remote teams, complete with video-chatting and messaging capabilities, as well as integrations with top enterprise tools, including Notion, GitHub and Trello.

[0] https://kwokchain.com/2019/08/16/the-arc-of-collaboration/

New comment by amsilprotag in "Whistle-blower reveals N Dakota leak of 11M gallons of gas condensate in 2015"

amsilprotag — Wed, 21 Aug 2019 17:32:03 +0000

From Robert Caro's The Power Broker:

Robert Moses had shifted the parkway south of Otto Kahn's estate, south of Winthrop's and Mills's estates, south of Stimson's and De Forest's. For men of wealth and influence, he had moved it more than three miles south of its original location. But James Roth possessed neither money nor influence. And for James Roth, Robert Moses would not move the parkway south even one tenth of a mile farther. For James Roth, Robert Moses would not move the parkway one foot.

New comment by amsilprotag in "Startups from Y Combinator’s S19 Demo Day 1"

amsilprotag — Tue, 20 Aug 2019 07:00:31 +0000

Only one crypto mention, and it's a company selling banks tech to detect crypto fraud.

TRM Labs: Banks are required to trace the source of their customers’ money. TRM helps banks identify and trace cryptocurrency fraud. They charge $20K per user seat. Though they couldn’t say the name, TRM says they recently signed a top-five global bank as a customer.

New comment by amsilprotag in "GitLab – A $1B business where all employees work remotely"

amsilprotag — Mon, 22 Jul 2019 18:00:15 +0000

I was pretty sure "say you wrote it" was an example in the original essay. Though upon review, Graham writes "almost verbatim".

Different publications vary greatly in their reliance on PR firms. At the bottom of the heap are the trade press, who make most of their money from advertising and would give the magazines away for free if advertisers would let them. [2] The average trade publication is a bunch of ads, glued together by just enough articles to make it look like a magazine. They're so desperate for "content" that some will print your press releases almost verbatim, if you take the trouble to write them to read like articles.

http://www.paulgraham.com/submarine.html

New comment by amsilprotag in "Does the news reflect what we die from?"

amsilprotag — Fri, 31 May 2019 01:16:41 +0000

The road death category is wrong. What is being reported is "accidents", of which CDC says there were 161,374 out of 2,744,248 deaths in 2016. But even this only gets you 5.88%, under their 7.68% reported in the interactive graph when hovering over 2016.

From [0]

Total Deaths 2016: 2,744,248

Unintended Injuries 2016: 161,374

From [1], a breakdown of those unintended 161,374 injuries above:

Accidental Poisoning and exposure to noxious substances: 58,335

Motor Vehicle: 40,327

Fall: 34,673

Accidental hanging, strangulation, and suffocation: 6,610

Accidental drowning and submersion: 3,786

Accidental exposure to smoke, fire and flames: 2,730

Dividing the 40,327 motor vehicle from the 2,744,248 deaths gets you 1.47%, way less than the 7.68 shown in the interactive chart.

[0] https://www.cdc.gov/nchs/data/hus/2017/019.pdf page 1

[1] https://www.cdc.gov/nchs/data/nvsr/nvsr67/nvsr67_05.pdf page 34

New comment by amsilprotag in "Ask HN: How do you stay disciplined in the long run?"

amsilprotag — Mon, 29 Apr 2019 19:29:20 +0000

I've noticed the same three categories discussed in two "go west" novels -- Steinbeck's East of Eden (1952) and Kesey's Sometimes a Great Notion (1962). It's interesting, and a bit jarring, to see the MBA translation of what, to me, existed as a literary aside.