a10zn8

[a10zn8 @a10zn8]

@igrekde хакатон длился 15 минут?

[Hasan Toor @hasantoxr]

🚨BREAKING: Microsoft Research + Salesforce just dropped a paper that should scare every AI builder.

They tested 15 top LLMs GPT-4.1, Gemini 2.5 Pro, Claude 3.7 Sonnet, o3, DeepSeek R1, Llama 4 across 200,000+ simulated conversations.

Single-turn prompt: 90% performance.
Multi-turn conversation: 65% performance.

Same model. Same task. Just... talking normally.

The culprit isn't intelligence. Aptitude only dropped 15%.

Unreliability EXPLODED by 112%.

→ LLMs answer before you finish explaining (wrong assumptions get baked in permanently)
→ They fall in love with their first wrong answer and build on it
→ They forget the middle of your conversation entirely
→ Longer responses introduce more assumptions = more errors

Even reasoning models failed. o3 and DeepSeek R1 performed just as badly.
Extra thinking tokens did nothing.

Setting temperature to 0? Still broken.

The fix right now: give your AI everything upfront in one message instead of back-and-forth.

Every benchmark you've seen was tested on single-turn prompts in perfect lab conditions.

Real conversations break every model on the market and nobody's talking about it.

[Robert Youssef @rryssf]

Microsoft Research and Salesforce analyzed 200,000+ AI conversations and found something the entire industry already suspected but nobody would say out loud.

every major model gets dramatically worse the longer you talk to it.

GPT-4, Claude, Gemini, Llama. all of them. no exceptions.

paper: https://t.co/W9KpYpIwui

[a10zn8 @a10zn8]

I have just claimed 5000.0 FLT tokens of Fluence! https://t.co/b8J8j6l7uf

[Seraphim @MacroMate8]

Quite a few protocols I’ve spoken to in Istanbul have switched partially or entirely to @drpcorg RPC service since it saves >50% in cost and is highly reliable

These guys don’t even have a token yet, pure focus on product

[dRPC // AI-powered RPC Infrastructure @dRPCorg]

Team @drpcorg is landing in Istanbul on Monday, ready to dive into @DevConnectIst. Meet our team of on #rpc #infrastructure experts, including @constantine_rm @thought_sync and @a10zn8. Reach out if you have any questions on #RPC in general or our platform in particular.

[a10zn8 @a10zn8]

Will join:
🔸DECENTRALIZED RPC SUMMIT
🔸HOT TAKE Series: Battle of the L2s
🔸Zero Knowledge Proofs Unleashed
🔸CHAINLINK ISTANBUL
🔸Abstract Istanbul
and more...
Can't wait to dive into these sessions & discuss the innovations in #dRPC with you all! 🚀 #CryptoConversations

[a10zn8 @a10zn8]

📢 Exciting news! I'm attending #DevConnect with #drpc team. Eager to engage with the brightest in #crypto and explore the latest in #decentralization. 🌟 Let's meet up & exchange ideas on the future of blockchain! #CryptoCommunity #BlockchainEvent

[a10zn8 @a10zn8]

@zk_chat 26d124add3f0cc129bcad1b34b7020da8109445f31e78d4363b7a7a4ffcd9536

[a10zn8 @a10zn8]

We're So Back

[a10zn8 @a10zn8]

@Golubtsova Но почему тогда нет секса 2?

[a10zn8 @a10zn8]

@k_moroznaya Положительная динамика разрушения личности?

[Vranac Srdjan @vranac]

YOU ARE IN A LEGACY CODEBASE
> RUN TESTS
YOU HAVE NO TESTS
> READ SPEC
YOU HAVE NO SPEC
> WRITE FIX
YOU ARE EATEN BY AN ELDER CODE HACK.

[Victor Grishchenko @gritzko]

Made a new Swarm RON implementation of CRDT Replicated Global Arrays. Use case: concurrently edited arrays/texts. Imagine git-like-Hackpad.

[Years Progress @YearsProgress]

▓▓▓▓▓▓▓▓▓░░░░░░ 60%

[a10zn8 @a10zn8]

@rexadecimal Как и любой другой

[a10zn8 @a10zn8]

@turbojedi а из нашего окна площадь инфракрасная видна

[🔎Julia Evans🔍 @b0rk]

bash tips (attached blog post: https://t.co/EPmt7BTP98)

[Daniel Peter @danieljpeter]

RT if you've ever typed AT commands like ATDT to control a modem