neit

prev pure math @YorkUniversity swe @SingleStoreDB @Google

[neit @ogintech]

i wish the AI's explanation will read more like a mathematics textbook: more intuition on why these approaches were tried & more natural progression. surely it would be at least more readable & more easily verifiable by human mathematician (or is it obfuscation as a feature?)

[Anthropic @AnthropicAI]

We asked an unreleased research version of Claude to take a stab at the Riemann hypothesis.

It didn’t solve it, but it did make strides on a related problem: it increased the lower bound for the fraction of zeros of the Riemann zeta function that satisfy the hypothesis from 41.6% to 67.2%.

https://t.co/aZDvqqhHRi

[neit @ogintech]

@QiaochuYuan it's interesting to finally see sb "formalize" this phenomenon. this has been rly bugging me because i want to learn things i might not aware of its existence yet and this defeates the whole purpose. curious how you deal w it tho.

[neit @ogintech]

@yminsky thought about this a lot but public models r mostly gud at coding py3 compared to hs/ml.
not many besides JaneS. in gud position to study type-enforced agent bc huge factor is in data quality that feeds the transformer. automate type-driven dev will be revolutionary too imo.

[neit @ogintech]

visiting the most gorgeous museum (easily mog the louvre) in cdmx & all i can think about is hosting a hackathon here. super spacious, peak coffee & gud value food nearby. only real concern is the art being so distracting that hackers stop shipping.

[neit @ogintech]

@0xTraderJoes beautifully written & a lot of things i wanted to say but wasnt able to express this well

[neit @ogintech]

funnily enough Google is the company that gives me the most number of interviews before offer bc of their team matching stage

interesting data tho, ig i should stop the process w a company if they reach > 4 interviews for being inefficient

[Jordan Mazer @justmazer]

Google's famous "Rule of Four" found that, at 4 total interviews, they could reach 86% confidence on a given hiring outcome - and each successive interview only added marginal utility (<1% confidence) while introducing considerable organizational tax (and dollar cost).

Two critical caveats about this finding...

- This was for well-defined jobs
- Administered by highly trained interviewers

If you're interviewing with a company doing >4 interviews, one of the following is likely at play:

1. They don't know what they're looking for (role clarity)
2. They don't know how to vet candidates (interviewing skill)
3. Hiring team "democratized" the hiring process (inmates run the asylum)
4. They don't trust each others' evaluations (org dysfunction)

... The first two are often the culprits in early stage companies. They really don't know what they are looking for, or how to evaluate it. Areas worth being nervous about, but definitely not "run for the hills" issues with a company.

The third - hiring managers who defer the decision to their team - is a ubiquitous problem in organizations large and small. If you are observing this behavior, just imagine the same indecision pervades all manner of other critical decision making (eg: feature priority, investment decisions, etc). This is a more serious flag than the first two issues.

The fourth is the worst of all of the possible contributors to meeting a dozen interview panelists. This version means there are likely warring silos within the organization, seemingly of equivalent might and heft, all of whom believe the other isn't capable of evaluating a potential hire. Of all the possible contributors to "way too many fucking interviews," this is the most concerning.

If you are interviewing with a company and you find yourself on your 7th, 8th, 9th, (N)th step... try to figure out which of these are occurring. It should tell you if this will be a somewhat annoying remediable situation when you join... or a catastrophic flaw that is unlikely to ever change.

[neit @ogintech]

@hanzpo don't let the streak fool you, half-assed lesson a day just for the streak &lt; focus speeding through sections for hrs a week, but still better than quitting entirely ig

[neit @ogintech]

@AaronAnandji so real, literally cancelled a hang-out with friends on 1st day of snow &amp; ended up locking in bc i got spooked by the wind &amp; snow when I stepped outside :(

[neit @ogintech]

@Sdavies22 wb 21-23 start so i can fly in &amp; out Vegas after playin w/o worrying about wkend hotel cost (book hotel + rest full day if make d2)
10 ok for 45m+ lvl, but 11-12 barely enuf to get up at 8-9, work out, food (most good food places not open until 11), also shower + commute time

[neit @ogintech]

forgot to update but my team won a prize (maybe partly due to the wifi not working but 2/4 of our phone data worked consistently xD)

[neit @ogintech]

sb must be hacking the hackathon's wifi for unfair advantage at @CalHacks bc how can it be this bad...

[neit @ogintech]

@0xTraderJoes a lot more could be put on the table for easy access :( inefficient use of space as a swe...

[neit @ogintech]

certainly cold it was, it rained: the flickering sound blopped against my basement window, almost chill like lofi beat. i crippled from my lack of sleep, trembling with cold as i watched 3 cpu x 96 cores maxing 100% capacity, exuding the delightful warmth i could only dream of...

[neit @ogintech]

@sahibdsv @HackTheNorth framework with the keyboard removed for cooling

[neit @ogintech]

someone is building their own os at @HackTheNorth with 5 laptops setup and im all for it

[neit @ogintech]

@budwindruid @HackTheNorth yes. dyk if he uses X?

[neit @ogintech]

@nizzyabi - will i get [email protected] after connecting through google? it's incredibly ironic when you guys want to beat gmail...
- ai feature in composing has been enough. - not sure wat u guys offered but im looking for ai optimization with labeling / creating forward rules

[Psyho @FakePsyho]

Let's make a silly experiment.
For every like this post receives in the next 24 hours, I will post one tip for heuristic/bot programming contests.

I will start posting these as a thread in ~1h from now.

I could talk about those things for days, so bring it on!

[neit @ogintech]

vibe coding is helping me save time, but the context switching window is so small that it's only enough time for me to reply to texts; optimistically in the long run, my dating life should still benefit greatly tks to ai (i believe)

[neit @ogintech]

if @duolingo is reading this, I also hope messaging will be a feature on the app. Would love to make friends with those who earn &gt;30k xp in a week in those hell diamond tournament :p

[neit @ogintech]

while mandarin is 2nd most spoken language (1.1 bil) &amp; 5th most L2 speakers (199 mil), it's not the most popular choice at any countries for almost 4 years. however, it's consistently the 2nd most popular at vietnam, lao &amp; cambodia for the past 4 years.
(more below)

[Duolingo @duolingo]

things you will learn if u check out the duolingo global language report: 2 truths and a lie

1) the most popular language studied on duolingo
2) the country that is 😐 not messing around 😐 with their lessons
3) my gym routine

(hint: i don’t go to the gym)

[neit @ogintech]

It makes me wonder if the popularity determines the investment in a language. For eg, both Spanish &amp; French have until B2 level on Duolingo, and for that reason, I've been spending most of my time on Duolingo for French (which wouldn't be the case if Mandarin content is better)