New Risk board released
HN user
T_S_
tim dot sears at pingwell.com
Mobile apps, computer vision, machine learning and Haskell. Our business combines all of these. I also have long experience in financial markets so I like to rant about the real meaning of financial transparency.
Not so sure about that. Check out https://github.com/ray-project/llmperf-leaderboard
And try mixtral on chat.groq.com
Anybody remember Kohoutek? Exactly. If I were a comet, my strategy would be to underpromise and overdeliver. That's the best way to rally a comet's soothsayer base.
It's Newtonian. People have no appreciation for how far you have come (position) or how fast you are going (speed), but they are extremely sensitive to acceleration. VCs especially so.
I am calling bullshit.
Amazing Fact (TM): I've been throwing right-handed for my whole life. Make me throw lefty and I throw like a girl. Something to do with an urgent need to get base runners out.
Another Amazing Fact (TM): If you play catch with an Aussie, notice their motion will be more overhanded than yours, if you are a Yank.
It's nurture not nature here. Go get a softball and try to fast pitch. I bet there's a girl nearby that could beat you.
Yeah, you can have a target market of $100 billion and not even crack 0.25.
I think I am at 0.25 on this scale.
The OP gives great advice. But I was amused by the Dorsey quote referring to "...founders - who may not have been there at inception."
Let's take some highly desirable attributes in an employee and give them a label "Rockstar". Er, no. "Founder".
When words are the currency of persuasion, expect inflation.
I have been using Haskell seriously for about two years now and humorously for a bit longer. For some reason I bought the old sales pitch some time ago. Now that I have a clue about the language I would like to rewrite the sales pitch for anyone considering a move.
Here's what happens in the middle of a big Haskell project.
1) Take some existing code.
2) Refactor (Add a feature, change a record element, etc.)
3) Run ghc, then fix all the errors it reports. Repeat this step as needed.
4) Done.
What would be different in python? For one, no compile errors means very little. How many test cases would you need? How would you find all the places the code must change?
Lesson: A tough, smart compiler can be the programmer's best friend.
Meta-lesson: There is little need to discuss laziness, purity or static typing to "sell" Haskell. They are enablers of ghc. What you need is a big, complicated code base and an urgent need to make changes to it. You won't have that until you get your feet wet.
Is Haskell perfect? No, but I'll save that for an OP that is overhyping the language.
Only as long as you use static linking.
As a corollary to this, keep your eyes on relative prices, incentives and human nature. Dull be highly effective.
What technology does is send some the prices of some actions from infinity to nearly zero. Meanwhile humans remain pretty much unchanged.
Politics is theater. At least under our current setup and probably most others.
Ever notice how little policy is actually discussed even when a supposed policy wonk (e.g. Clinton, Ryan) are giving the speech?
Why can't the TSA make me a suit? They already have the measurements. I'm barely half joking here. Seems like this is one of those unsolved "solved problems".
Reminds me of my kid's $6000 bump on the head. Or the time I had and knee operation and they took five minutes extra to clean out something they spotted and added $3500 to the bill. Or the eye operation where they took fifteen minutes extra because there was a detached retina adding $9000 to the bill. (Insurance wouldn't pay that one. They never said why either.)
You are not exactly going to negotiate at moments like these.
Not much content in this blog piece. Couple of isolated quotes from a Forbes article from a with no tl;dr on offer.
Instead of "Capitalism is...beautiful" ideology it would be worth focussing on when it works and when it doesn't. Seems that there is strong evidence that capitalism harness human nature, property rights, and incentives to enable production. That's great. There is a lot less discussion about what it takes for markets to allocate resources efficiently. Two of the bigger prerequisites are "no externalities" and "symmetric information". Capitalism has no built-in mechanism to establish these conditions. A lot of our problems today go right to those factors. We could do something about it if our politicians would stop playing to the cheap seats. Sorry capitalism, you're ok but not that beautiful.
A little story about the jury system. Years ago I was a juror on a murder trial in NYC. The judge was strict beyond belief with the attorneys. He seemed tougher than any hanging judge in a movie, though not unfair. The verdict was reached quickly: guilty. Every one associated with the trial went on with our lives. All but the victim and the defendant.
A year later, the judge retired and published his memoir. A big surprise awaited. He had begun his career as an ACLU lawyer in the 60s but wound up with a tough reputation among the defense bar.
An even bigger surprise awaited me. He devoted a chapter to explaining how hopeless the jury system is. To illustrate, he told the story of a murder case. My murder case. I read, "Never was a defendant more plainly guilty". I agreed, with some relief, but was confused. Why this case?
Turns out my trial was the second trial. At the first, some of the female jurors refused to convict. The defendant was "too good looking to have committed the crime", in their opinion. Mistrial. Then I recalled how, in our later trial, two female jurors wanted to find a way to justify a manslaughter conviction for what was clearly deliberate murder. They felt sorry for the handsome young defendant.
You be the judge (or jury), but if I had not lived it, I would not have believed this could happen.
If you make me grade my relationships with people (friend/defriend/link/put in circle etc.) you are making me work. You must reward me for that. On all these social networks my reward for that exercise is actually some negative feelings, netted against the positives of getting my message out properly. Maybe that's one reason Twitter is so popular among some, you don't really need to drop followers, you just tweet.
Nice article by NYTimes. Too bad about the linkbait headline, but, hey, they have to sell clicks too. At Hacker Dojo, we're working hard to raise the money so everyone can just get back to writing code, safely egressing etc.
This is not confined to food. In my house probably 40% of cheap imports end up in the trash in a year or two. During WWII, fully half the U.S. economy was devoted to the war effort. Nobody starved, they just drove on bald tires and stayed in the same house. We need a carbon tax (or a consumption tax, for the climate change deniers) and this sort of thing would disappear.
Someday through historical echo transcription (YC14 perhaps), the real transcript will emerge...
Me: Hi Steve, remember how I asked all the managers to me with their direct reports 1-on-1 at least once every six months.
Steve: No, when was that?
Me: Never mind. I heard that Tim didn't do that.
Steve: Do what?
Me: Meet with his direct reports. 1-on-1.
Steve: Oh yeah. Reminds me, great Celtics game last night, huh?
Me: I tivo'd it and caught the last 5 minutes live. Saves a shitload of time wasted on commercials. Cable is gonna die a painful death. What were we talking about? Oh yeah. Please tell Tim to follow up or I will be pissed.
Steve: Sure you don't want to tell him yourself? He never listens to me.
Me: OK, maybe tomorrow. Remind me in 24 hours.
"Embrace and extend." -- Bill Gates
These days that would correspond to forking the spec. So unlike an open source programming, it is not nice to fork the spec.
Or is it?
Beg to differ.
Whenever I walk through airport security, I have Osama to thank for the millimeter wave inspection. And for the pain of raising money from non-accredited investors I have hundreds of historical and would-be scammers to thank. Bad guys make us limit our freedom.
Although we ask the government to protect us, they usually mess it up by ham-handed regulation. At the airport, the big show makes us feel safer, even though some experts say we aren't. The same thing has been going on in finance for years with Congress and regulators asleep at the switch with no idea how to control the bankers. They lack imagination and incentives.
We need to completely rethink regulation for investment vehicles. The goal should be to improve post-issuance transparency. Real transparency.
First thing we need is a live feed of transactions logs. All of them. Dirt cheap these days. No fancy accounting, I'll model the company myself thanks. I can share models with fellow investors. I/you/we can sell modeling capability to investors. More sunlight--less corruption, less risk.
The lack of transparency and disclosure that may about to be enshrined in Regulation A is like handing out the box cutters on the airplane to al Qaeda. The threat is not you or me. It's scammers. Right now about all the protection Mom will have is the SEC banning "bad actors" years later--if they get caught. Just like they caught Madoff. Professional financial accounting? Enron's books were audited by a top firm.
Instead regulation should require you to show me the money, continuously, with timestamps, please. Oh, and send me my government mandated Salesforce.com login too please. Let's do for banks too. Make it fair for everyone.
OK, now you can sell my Mom your stock.
Maybe Fidelity? PIMCO? Humble Warren B. of Omaha. Lots of other examples. Money is universal...like love. Kind of.
I took it as more of a nod to the old emacs joke (probably told by a vim user): “Emacs is a great operating system; too bad it doesn’t have a decent editor”
The title is linkbait if not trollish. However the article is just a technical discussion about the best way to implement exceptions in pure functional code. He prefers the way it is done in ML.
Really great. I wonder if there are other algos that are being overlooked because we are looking at them with the wrong cost model in mind.
Math is particularly tough on the ego. I remember one of my advisors said: "I try to decide if something is true. I work extremely hard to do so. Then when I am done, it seems clear that it was true all along, and the only problem was that I didn't know it."
Max 2. Local courts are "state" offices.
OK I'll bite. What's your number?
Let's see. 1 generation => Lots of compassion. 6 billion years => no compassion. So somewhere in between. How about you?