Explaining this will require setting up a bit of terminology. I'm not going to comment on Telegram's security because I don't want to get involved in a Signal-T

Explaining this will require setting up a bit of terminology. I'm not going to comment on Telegram's security because I don't want to get involved in a Signal-Telegram flame war. But I can at least explain why the contest is a poor methodology for ascertaining the security of a cryptosystem. To begin, when researchers publish or analyze cryptographic algorithms, they model the interaction of legitimate users and adversaries. Two legitimate users will engage in communication and an adversary will sit "in the middle", trying to break ciphertext confidentiality or authenticity (or both). Generally the adversary is allowed to tamper with or reply messages from Alice to Bob and vice versa. Adversaries are alsl given various capabilities of interacting with the cryptosystem and messages therein, and if the cryptosystem is secure against an ever more capable adversary, we say it is "____-secure." Researchers model the interactions in this way to more closely approximate real world, repetitive usage instead of an isolated simulacra. So let's talk about cryptanalytic attacks. There are known plaintext, chosen plaintext, known ciphertext and chosen ciphertext attacks. The attack always has a ciphertext, so known ciphertext is the lowest form (least assured) level of cryptanalysis. A known plaintext attack gives the attacker a corresponding ciphertext and a plaintext. A chosen plaintext attack affords the attacker the ability to use an encryption oracle to encrypt arbitrary plaintexts and see the ciphertexts. And likewise, a chosen ciphertext (the most powerful attack) allows an adversary to decrypt arbitrary ciphertexts. The Telegram contest requires you to provide the plaintext corresponding to a single ciphertext. You are not given access to real time encryption or decryption within a channel and you are not provided with the ability to attempt cryptanalysis on numerous messages. You can't even tamper with messages or try to replay them, because you have an isolated ciphertext to examine. Everything else is outside the scope of the contest. That means the only thing you're allowed to do is a known ciphertext attack. A known ciphertext attack is extremely difficult to pull off even for weak cryptography and it does not resemble the structure of a real world attack. For example, legitimate attacks on TLS are not known ciphertext attacks. That's why the Telegram contest is synthetic. Again, I'm not going to comment on its actual security. But if you want to do real cryptanalysis, you hire a team of cryptographers who didn't design the thing and try to break it with a more generous suite of capabilities that better resemble real world scenarios. The critiques from the cryptographic community should be examined on their own merits, and not judged according to whether or not they've successfully broken something under a known ciphertext attack.