piyush @Piyuple

swe, ai github.com/piyuple Joined February 2011

Tweets

11K
Followers

553
Following

650
Likes

6K

piyush @Piyuple

2 years ago

@steren Faced something similar this week and the fix was pip install -U packaging

0 0 0 9 0

View Details

maharshi @maharshii

2 years ago

link: lmsys.org/blog/2024-02-0…

0 3 46 17K 60

View Details

To summarize: •2’s complement is used to represent negative numbers in binary. •MSB = sign bit. •Range: -2^(N-1) to 2^(N-1) - 1. •Simple: invert + add 1 for negatives. •Arithmetic is seamless! It’s elegant and foundational to computing!

0 0 0 38 0

View Details

piyush @Piyuple

2 years ago

Why is 2’s complement so powerful? Arithmetic just works. E.g., 5 + (-3) in 4 bits: •5 = 0101 •-3 = 1101 (2’s complement of 3) •Add: 0101 + 1101 = 10010. Drop the carry bit: 0010 → 2. No special logic needed!

1 0 0 36 0

View Details

piyush @Piyuple

2 years ago

It also solves the +0 and -0 problem found in 1’s complement by having a single representation for 0: 0000.

1 0 0 26 0

View Details

piyush @Piyuple

2 years ago

Let’s talk about 2’s complement, one of the most important concepts in binary arithmetic and computer science! It’s how computers represent negative numbers and perform seamless arithmetic. 🧵

1 0 0 40 0

View Details

piyush @Piyuple

2 years ago

In 4-bit 2’s complement, negative numbers wrap around the circle starting at 2^4 = 16. •-1 → 1111 (decimal: 15 in unsigned). •-8 → 1000. Range: -8 to 7. This allows seamless addition/subtraction without extra rules!

1 0 0 26 0

View Details

piyush @Piyuple

2 years ago

How do we convert a 2’s complement binary back to decimal? E.g., 1011 in 4 bits: 1️⃣ Check MSB: 1 → Negative. 2️⃣ Take 2’s complement: •Invert: 1011 → 0100. •Add 1: 0100 + 1 = 0101. 3️⃣ Result: -5. For positives, MSB is 0—no conversion needed!

1 0 0 24 0

View Details

piyush @Piyuple

2 years ago

Why is it called 2’s complement? Because subtracting a number is like adding its complement relative to 2^N. E.g., for 5 in 4 bits, 2^4 = 16: •Complement = 16 - 5 = 11 (binary: 1011). This clever trick simplifies arithmetic in binary!

1 0 0 20 0

View Details

piyush @Piyuple

2 years ago

To find the 2’s complement (binary representation of a negative number): 1️⃣ Write the positive number in binary. 2️⃣ Invert all bits (1’s complement). 3️⃣ Add 1 to the result. E.g., for -5 in 4 bits: 1.5 = 0101 2.Invert: 1010 3.Add 1: 1010 + 1 = 1011.

1 0 0 18 0

View Details

piyush @Piyuple

2 years ago

In N-bit 2’s complement, the MSB (most significant bit) acts as the sign bit: •0 → Positive •1 → Negative. Range: -2^(N-1) to 2^(N-1) - 1. E.g., in 4 bits: •Min: -8 (1000) •Max: 7 (0111). Let’s break this down further!

1 0 0 19 0

View Details

piyush @Piyuple

2 years ago

Why do we need 2’s complement? Computers work with binary (0s and 1s). To handle negative numbers in binary, we need a system that: 1️⃣ Represents negatives uniquely. 2️⃣ Simplifies addition/subtraction. 2’s complement solves both efficiently.

1 0 0 17 0

View Details

yan @bcrypt

2 years ago

reminder that the bcrypt hash function ignores input above a certain length! so if you do bcrypt(username || password) for some reason, a sufficiently long username will make it accept any password. to fix this you can sha256 the input first.

Kinnaird McQuade @kmcquade3

2 years ago

Okta allowing login bypass for any usernames with 52+ characters is insane Official Security Advisory: trust.okta.com/security-advis…

35 365 2K 1.0M 441

62 988 6K 748K 2K

View Details

ThePrimeagen @ThePrimeagen

2 years ago

the biggest trick the devil ever played was convincing management that MBAs should determine how software planning works

72 162 2K 101K 121

View Details

piyush @Piyuple

2 years ago

@ThePrimeagen I locked myself out of a new ssh only device once by enabling ufw and not allowing port 22.

0 0 0 163 0

View Details

merve @mervenoyann

2 years ago

solving problems using LLMs that can be solved by fine-tuning BERT is a skill issue

111 246 3K 642K 1K

View Details

NeetCode @neetcode1

2 years ago

I was just watching a video about algorithm performance by the @ThePrimeagen . It reminded that Big-O complexity is not always the final story. When choosing a data structure, it's useful to consider the 'composition hierarchy'. You can implement an algorithm using a Hash Table. In theory, it provides O(1) insert & lookups. But don't forget that a hash table is just implemented using a dynamic array. With small amounts of data, a linear scan through an array is faster than hash lookups. And if you have a fixed amount of data, why not just use a static array? (prob because you use a fake language like JS or Python) Take it one step further. If your data can be binary encoded, you can get further performance gains. Possibly at the cost of readability. Stepping back, you realize that this is the same principal that Protocol Buffers are built on. Anyways, I love thinking about this stuff and thought it was worth sharing.