statusgraph

That seems... fine? The terms basically imply that if you publish a benchmark you need to let CRDB reproduce your benchmark and discuss it publicly

It's comical how much of modern startup compensation is gymnastics around taxes.

My understanding is that the most basic problem is if the startup gives you stock that's viewed as income and you will need to pay tax on an illiquid asset, and ISO/NSO/RSU/409a's are resulting from various elaborate schemes to give you stock-like upside tomorrow without having to pay taxes on an illiquid today.

Bard has a setting to enable something approximating streaming responses (still not quite as nice as GPT)

Seems like it, the parent poster is almost implying there's an element of underwriting the transaction/house for which 2% is not unreasonable (Title insurance is in the 1% (or a bit less) range in the US I believe)

Yeah, it looks like it's pro-rated:

The Settlement Administrator will assign each Authorized Claimant one point for each month in which the Authorized Claimant had an activated Facebook account during the Class Period.

(#8 on https://www.facebookuserprivacysettlement.com/#faqs )

This is fascinating, but is focused on the engineering side and not the modeling: https://www.youtube.com/watch?v=b1e4t2k2KJY

Just be mindful of the "Unauthorized Visitor Fee" --

“Whether or not related to Consumertronics business, any visit to our facilities which is not with our clear, written, prior permission, or so permitted by us based on a fraudulent, erroneous or deceptive representation by the visitor, is an unauthorized visit which automatically results in a non-refundable Unauthorized Visitor Fee (UVF) for each visitor and for each visit for which the visitor assumes total personal liability for paying. The amounts of the UVFs are set by us based on our perception of the real and potential injury to us caused by the visitor and visit, as well as the amount of our time and inconvenience for the unauthorized visit. The UVF amounts may vary between $0 and $1,000,000+ per visitor per visit. UVF fees are independent and separate from all other fees, charges and costs the unauthorized visitor may be required to pay.”

Jokes aside Comic Mono is surprisingly nice. I thought I'd hate it but it has a fun whimsy to it

By almost 5% actually! https://www.google.com/search?q=11hr+22min+%2F+11hr+55min&oq...

Well structurally taking Twitter private has increased expenses by $1bil (the interest on loans) against last year's ~$6bil in revenue, so that alone is going to change the health of the company.

On top of that Musk has given guidance that he expects revenue to fall from $6b in 2022 to $3b in 2023. (This seems dramatic TBH)

I don't think there were revelations (this is all public info), the transaction just changed the fundamentals.

Company hackathons are rarely 24hr affairs and more "don't do your normal work during working hours"

Argh, thanks for flagging. Looks like I can't edit :(

Yeah, I absolutely agree with a common tactic being dangling an ML carrot at recruiting but the work not even being ML related.

I've historically seen the ML team be separate from Data Science/Analytics, I wonder if that helps with this

BQ Slots lets you do essentially that (pre-commit to a particular cluster size)

Two-phase commit deals with network failure, not change semantics. The message itself (whether "BALANCE=40" or "SUBTRACT 40") isn't something that matters to 2pc.

I've found the Yahoo paper from 2010 to be the best distillation of the general problem of feed following and how to solve it (blending push/pull as you say):http://jeffterrace.com/docs/feeding-frenzy-sigmod10-web.pdf

We tried spinnaker and are now using Weave (https://www.weave.works/). It works pretty well, actually. In particular we like that it stores all state in a git repo which is automatically updated.

We have our own monitoring stack, though, so don't use the additional prometheus integrations

No, for the same reason you don't spend most of your time on your weakest employees or most of your money on your poorest investments. Feed your winners.

Citizen | New York, NY | Full time | Onsite www.citizen.com

We are building a global safety network. The need has never been greater for technology that informs and protects the public. Citizen empowers communities to keep each other safe and informed. Citizen is backed by Sequoia Capital and Founder’s Fund, and is a top 10 global News App on the App Store.

We’re looking for passionate, hardworking, mission-driven people to help take the Citizen app to the next level and launch multiple cities in 2018. Our New York City office is the center of our operation, and the hub of all engineering and product activities.

---

Some of the problem spaces we work in today:

Scaling: We have one of the fastest growing organic user bases in NYC, and right now we’re geo-restricted to New York City and San Francisco. Soon we will launch nation-wide and we need to support that scale. A Citizen international launch will follow.

Bursting: Without warning, a major incident occurs and multiple users begin live-streaming video into the app. We notify the entire user base, causing a massive burst of simultaneously connected clients. The core systems we build need to efficiently support traffic that can grow to hundreds of thousands of concurrent users today, and millions in the future.

---

We're hiring across all engineering disciplines. Send me an email to wkm@citizen.com!

Yes, this has been used as a defense against insider trading accusations, sometimes unsuccessfully: https://en.m.wikipedia.org/wiki/Mosaic_theory_(investments)

Interestingly, you can treat the NN as a black box (ie, not look at individual weights or even the architecture) and still derive adversarial cases:

https://arxiv.org/abs/1602.02697

You can. Tensorflow has a CPU implementation. But, it's not a little longer, it's incredibly longer.

I get the same response when I ask "does my insurance work outside of new york state". It doesn't seem any more intelligent than finding keywords.

This is precisely what is happening in Somalia (as noted in the article), although in the inverse: the value of the money fluctuates (down) with the cost of production.

Most likely you can only breed sufficiently healthy animals.

Note that Kafka only guarantees message ordering per-partition

Not quite. Big-O creates an -upper bound-. It could be the upper bound of the best case, the average, or the worst case. (So, for example, one can say quicksort has worst case performance O(n^2) but an expected performance of O(n log n))

How interesting! I believed too much of the A7 stuff =)

This may be the future, but recall that the GPUs in Macs are not made by Apple or exclusive to their platform in the way they are for the iDevices.

Ctrl + left or right arrow