fallmonkey

Appreciated the response! I noticed the same when I ran tau2 myself on gpt-5 and 4.1, where gpt-5 is really good at looking at tool results and interleaving those with thinking, while 4.1/o3 struggles to decide the proper next tool to use even with thinking. To some extent, gpt-5 is too good at figuring out the right tool to use in one go. Amazing progress.

Correct, and with reasoning, the ratio is totally off. As others have pointed out, actual usage is way higher (much more than 3-5x) than the estimation in the article, which is probably for very trivial users.

The estimation for output token is too low since one reasoning-enabled response can burn through thousands of output tokens. Also low for input tokens since in actual use there're many context (memory, agents.md, rules, etc) included nowadays.

There’s an interesting dynamic here. What value of OpenAI to use for conversion calculation? 86b round is pretty much dead if they move to MS, yet 29b is too low (even 86b is low in terms of future potential). And what kind of upward room there will be?

I have no doubt that MS can spend billions in cash or RSU to compensate all of them, but I do believe there’s some gotcha if exodus actually happens and MS might not be so generous in throwing millions of dollar cash for a general backend engineer recently joining OpenAI.

On personal anecdotes, I've used the same internet identity across all platforms and it's not hard to locate my real life identity from there...since like 20 years ago when I got on the internet. So all the good reputation and bad reputation followed me through, and surprised me in best and worst way.

There're times when I wished I had concealed myself saying certain things (or just haven't said it), and times when I felt the joy of recognition when people connected dots about me and became closer friend. Now as I'm getting old, it's just a habit, no longer about risks or rewards.

Yet, both younger me and older me will probably be very indecisive about doing this if they've never done it and are told about all the fortune/misfortune this has brought to me, including meeting my wife and thinking about suicides.

"A common question I get is about work/life balance. And I view it a little different too now. At Facebook I could get by barely working a few hours a day. But the job was so unsatisfying that it spilled the frustration in the “life” part"

This is the golden piece for me. I too recently left a big corp to a smaller place, expecting to sacrifice WLB and my free time - yet I ended up feeling happier and spending less time consuming short-form online contents, to the point that I even have more free time now! It never occurred to me that my self-indulging online addiction is actually a coping mechanism towards the unsatisfying and meaningless daily job.

Yet you might be surprised to know that, in China, all those "offline games" are categorized as online games because they all can be distributed online, when it comes to import reviews. No game is safe from this, really.

With Google doing front loading(33/33/22/11) and other companies also shifting away from conventional 25% per year vesting to make first few years more attractive, they gotta catch up to stay competitive. Uber recently did the same of removing cliff.

I'd like to provide an personal take on this - this will be a bad deal until Uber can be a super app doing 100x more things than it's doing today. So maybe never.

Why? Because exactly in China there was consolidation of food delivery apps into like 3 or 4 of them from 100+. But all of them are still burning lots of money from investors, except one - Meituan-Dianpin which built on this super app (first think of it as uber+yelp+groupon+tripadvisor+more) handling all kinds of life needs from food to labor service. Yet this super app barely started making profits recently after grabbing so much aspects of daily life, while its food delivery unit is still burning money, though contributing to the profiting units like ads.