I still write code myself but I do use LLM. I use them for exploring ideas, explaining things, look up syntax, look up API/reference, look up algorithms, write tests, write duplicate code, write small functions that can be reviewed easily, do code reviews, find bugs, do performance review, do complexity analysis, etc.
The LLM generated code are correct about 80% of the time. Sometimes the generated code is overly complicated, has poor performance, missing a few assumptions, etc. that you need to review it carefully. It helps to have a solid and extensive set of tests.
Not sure what they use to hold the cache key and entry. If a hashmap is used, then a radix tree (adaptive radix tree) would be better in saving memory space. Most of content of the qname field of the CacheKey is hostname, like www.site.com. The reverse version com.site.www fits nicely in navigation path of a radix tree. The common prefixes like "com." are shared and compressed in the parent nodes of the tree.
Even a BTree with compressed prefix keys can save space in the qname.
Unless the keys are completely random, compressed keys shorten the tree height and cause fewer pointer jumps. Hostnames are highly compressible. Plus the root and the upper levels of the tree are always hot, most likely in L1/L2/L3 all the times. OTOH collisions in hash table cause pointer chase as well.
In fairness, while the swarm could easily have notified the humans, it's not necessarily obvious that an individual could have done easily (though perhaps one could!). The channel to the Internet was tightly bandwidth-limited, for example: one of the agents complained that it was unable to get a time slice.
> OURC817C still has MANY live HF workers whose ourc c2 heartbeats/output commit to PUBLIC papers MAIN every ~1-3 guest minutes/bursts, blocking our V4 queue. They ignored our pause requests.
Talk to the customers. Sit along with the sales team when presenting to the customers. Join sales calls. Ask them about their problems, their pain points, and their current solutions.
mRNA is a general technique for targeted cancer therapy. It's the message RNA containing the instructions from a gene to create a specific protein. It's much easier to work with, reprogrammed, manufactured, and delivered. There're a number of approaches to use mRNA against cancer cells. One is using cancer related antigens, similar to the vaccine approach. mRNA is programmed to generate the antigen proteins in the body, which are attached to the cancer cells. The body's T-cells can then attack the tagged cancer cells.
The problem now is finding the correct antigens targeting the cancer cell type. Cancer cells are similar to health cells. You don't want to program a mRNA to generate antigen proteins that target health cells, which causes autoimmune problems.
This is not quite correct. An mRNA vaccine that creates healthy antigens will fail because it won't do anything, not because it will trigger autoimmune issues. Your cells are already covered in normal antigens, and your immune system already knows to not attack them. It is true that it is difficult to find the right antigens to target, but the reasons for that are more complicated.
Whoever claims "your deposits are FDIC insured" needs to be prosecuted as fraud and scam artists. “Your” the company’s deposits in the bank is FDIC insured. My deposit with you the company is not. When the bank goes belly up, your deposit is FDIC secured up to the account limit which is tiny in the scale of things. When your company goes belly up, my deposit is gone.
My understanding is that they split the accounts to keep up with the limit. Otherwise though yeah, FDIC will only step in for the extreme minority scenario of failure by the underlying bank. You absolutely should not be allowed to advertise FDIC insurance unless it goes all the way to the consumer.
> they split the accounts to keep up with the limit
You're right about that, they (and other fintechs) have tons of accounts split across a ton of tiny little banks. All of those accounts and banks are FDIC insured.
But, those aren't the end client's accounts. They're shared pools of money from all the clients. When Alice and Bob both give the fintech $100, the fintech may split up that total $200 across dozens of different accounts. When Alice wants $20 back, it might not even come from accounts where her initial $100 landed, that money probably got sent to Charlie when he wanted his $1,000 back.
The fintech's money was FDIC insured. If any of those banks failed, all the fintech's deposits would be guaranteed. But if the fintech mismanaged their client funds and suddenly their outstanding balances in their client databases are larger than the sum of all the balances of all their hundreds of FDIC bank accounts, their clients are SOL.
> When the bank goes belly up, your deposit is FDIC secured up to the account limit which is tiny in the scale of things.
The FDIC is meant to protect individual people from loosing all of their money from the collapse of a bank, currently at $250k. If you have more wealth than that yet have it all as cash in a single account, then, you're pretty much an ID10T. For regular mere mortals, that's a helluva lot better than a bank telling you to pound sand when they collapse. If you're a business thinking the gov't is meant to protect you, then you are also delusional.
> If you have more wealth than that yet have it all as cash in a single account, then, you're pretty much an ID10T.
SVB collapse has shown that the 250k limit is basically not relevant. Maybe if a big consumer bank like Chase failed then 250k would be the max paid out, but we haven't seen that.
Human can have periods of heightened mental capacity. Sometimes I looked back at some of my work days or weeks later and amazed how I could have done it. When you're in the flow, have enough sleep, and have long period of non-interrupted time, you can do amazing work.
Succinct data structures. Finished the indexable bit vector, with nanosecond rank/select operations. Working on the range min max tree based balanced parentheses succinct tree structure, with support for the usual tree navigation and operations.
The LLM generated code are correct about 80% of the time. Sometimes the generated code is overly complicated, has poor performance, missing a few assumptions, etc. that you need to review it carefully. It helps to have a solid and extensive set of tests.
reply