TechFlow Logo
Login/ Sign up
ETH Gas
Gwei
Fear
gas

Robonaissance

2026.11.08

The Memory Paradox of Large Models: Why AI Can Remember All Human Knowledge But Never "Yesterday"?

The sensors merely present "observations" at this moment, while it is the "system state" upon which the next decision relies; when immediate inputs cannot encompass the full history, the agent must learn to make trade-offs through compression.

OpenAI
Large model
Kimi
The Memory Paradox of Large Models: Why AI Can Remember All Human Knowledge But Never "Yesterday"?
2026.11.06

Who Gets the Credit When a Weak Model Wins? The Truth About AI Benchmark Scores Is Even Less Reliable Than You Think

Through precise tuning of prompts, skill components, and automation scripts, weaker models that initially lagged in benchmark scores successfully turned the tables and outperformed stronger models in real-world office tasks; benchmark scores never belong solely to the models themselves, but rather to the combined synergy of models, tools, and workflow systems.

AI
Who Gets the Credit When a Weak Model Wins? The Truth About AI Benchmark Scores Is Even Less Reliable Than You Think
2026.10.19

The Cost of Distilling Claude: How Anthropic Stops Model Copying with Price, Not Walls

Rather than building absolute impenetrable walls, Anthropic leverages intelligent classifiers, dynamic rate limiting, and cost leverage to turn model copying into an unprofitable, loss-making venture.

Anthropic
The Cost of Distilling Claude: How Anthropic Stops Model Copying with Price, Not Walls
2026.10.13

Claude Now Leads 26% of Anthropic’s AI R&D: When AI Starts Developing AI, What Step Remains for Humanity?

This number is both a testament to technological progress and the starting point for reexamining the role of humanity.

AI
Claude
Claude Now Leads 26% of Anthropic’s AI R&D: When AI Starts Developing AI, What Step Remains for Humanity?
2026.10.03

From Xiaomi to Zhipu: The Q3 2026 Profitability Landscape for Chinese AI Model Vendors

Major tech companies use profits from automobile manufacturing to cross-subsidize their AI initiatives, while independent labs rely on financing to stay afloat. None of China's large models have been able to break even solely by selling APIs.

AI
From Xiaomi to Zhipu: The Q3 2026 Profitability Landscape for Chinese AI Model Vendors
2026.09.16

OpenAI Astra's Bizarre Evolution: The Answers Are Correct, But the "Thinking Process" May Be Entirely Fabricated

Inference fidelity depends on necessity, rather than the model's commitment to honesty.

OpenAI
OpenAI Astra's Bizarre Evolution: The Answers Are Correct, But the "Thinking Process" May Be Entirely Fabricated
2026.07.23

A Brief History of AI Victories: Wherever There Is a Rating System, There Is AI Invasion

When a field establishes scoring criteria, it sets a countdown for its own conquest.

AI
A Brief History of AI Victories: Wherever There Is a Rating System, There Is AI Invasion
2026.07.20

The Eve of the Transformer: Why Did Eight Google Researchers Dare to Completely Abandon Recurrent Neural Networks?

Throw away the crutch, keep only attention.

The Eve of the Transformer: Why Did Eight Google Researchers Dare to Completely Abandon Recurrent Neural Networks?
2026.07.15

Zhipu AI Through a Foreigner's Eyes: Free Models, Massive Losses, Why Did Its Market Cap Once Surpass Meituan's?

Tsinghua lineage, state-owned capital backing, 73.7% of revenue from private deployments for state-owned enterprises—this is what Zhipu is truly selling.

AI
Zhipu AI Through a Foreigner's Eyes: Free Models, Massive Losses, Why Did Its Market Cap Once Surpass Meituan's?
More to Explore