All videos
0:00 / 0:00
research

The misaligned incentives behind AI coding agents

LangChain31 July 2026Watch on YouTube

What you'll learn

  • Devin is an AI software engineer that has evolved: when the first version achieved 13% on SWE-bench, Devin later became the number one contributor to its own codebase.
  • Smart model routing and selecting the right model (Fable vs GPT) can improve cost-performance by 35% without cutting quality.
  • AI agents become effective when wired to be proactive rather than reactive, enabling them to take independent initiative.
  • Mergeability of merged code becomes the next evaluation bar, measuring whether agent-written code truly integrates into existing systems.
  • Cognition backs a $10 million productivity guarantee and scores sessions as productive or unproductive through clear metrics.

Frequently asked questions

What is the difference between using Fable and GPT models in AI agents?
Fable and GPT models have different strengths: you choose based on the task and the economic trade-off between speed, cost, and quality. Smart routing between these models can deliver 35% better price-performance.
Why is letting users pick their own model considered a UX bug?
When users manually select models, they often make suboptimal choices. Intelligent routing in the background automatically picks the best model per task, delivering better results and lower costs.
What does 'intelligence saturation' mean in the context of AI coding?
Intelligence saturation occurs when engineering work becomes complex enough that speed and cost matter more than raw intelligence. This happens when AI agents are capable enough and companies shift focus to efficiency.
How does Cognition measure whether an AI agent session is productive?
Cognition uses clear metrics to score sessions as productive or unproductive. This enables them to underwrite a $10 million productivity guarantee and provide customers with insight into real value.

Topics

ai agentscoding agentsdevinmodel optimizationai economics

Read next

Description from the channel

Russell Kaplan is the President of Cognition, the $26 billion company behind Devin, the AI software engineer. He started his career in machine learning at Tesla's Autopilot team, then led the ML org at Scale AI, which had acquired his computer vision startup Helia. In this conversation, Russell unpacks why more and more engineering work is becoming "intelligence saturated," how the sidekick architecture inside Devin Fusion cuts cost without cutting quality, and why Cognition is underwriting a $10 million productivity guarantee. We also discuss: • Why mergeability is the next eval bar • When to use Fable vs GPT models • How smarter routing unlocked 35% better price performance • Why letting users pick their own model is a UX bug • Wiring agents to be proactive instead of reactive • How Cognition scores a session productive or unproductive Timestamps: 00:00 Introduction 01:21 Launching Devin at 13% on SWE-bench 02:47 When Devin became the number one committer to Devin 05:56 Intelligence saturation: when speed and cost start to matter more 07:05 When to use Fable vs GPT models 08:41 Why mergeability is the next eval bar 10:51 How open source maintainers shaped FrontierCode 14:56 The telltale sign of agent-written code 17:56 When per-person token spend starts to eclipse salaries 20:26 Why letting users pick their own model is a UX bug 21:46 How smarter routing unlocked 35% better price performance 23:50 Why Cognition still trains its own models 27:13 Deploying Cerebras at scale at 950 tokens per second 28:21 The Tesla chip debate: does an inference chip need division 30:06 Wiring agents to be proactive instead of reactive 32:19 Why "software factory" is the wrong term 35:16 Why everyone is now a new grad 41:10 Forward deployed engineering, and why ex-founders thrive 43:16 The industry's incentive to get you tokenmaxxing 45:11 How Cognition scores a session productive or unproductive 46:51 Underwriting a $10 million productivity guarantee 48:11 Using productivity dashboards to coach, not police References: • Anthropic: https://www.anthropic.com/ • Cerebras: https://www.cerebras.ai/ • Claude Fable: https://www.anthropic.com/claude/fable • Claude Opus: https://www.anthropic.com/claude/opus • Cognition: https://cognition.com/ • Cognition's AI Productivity Guarantee: https://cognition.com/blog/ai-guarantee • Devin: https://devin.ai/ • Devin Desktop: https://cognition.com/blog/introducing-devin-desktop • Devin Fusion: https://cognition.com/blog/devin-fusion • Devin Review: https://cognition.com/blog/devin-review • Devin Security Swarm: https://cognition.com/blog/introducing-devin-security-swarm • Estimating the Productivity of an Autonomous AI Software Engineer: https://cognition.com/blog/ai-productivity • FrontierCode: https://cognition.com/frontiercode • GPT-5.5: https://openai.com/index/introducing-gpt-5-5/ • GPT-5.5-Cyber: https://openai.com/index/gpt-5-5-with-trusted-access-for-cyber/ • Harbor: https://www.harborframework.com/ • LangSmith: https://smith.langchain.com/ • OpenAI: https://openai.com/ • OpenRouter Model Fusion: https://openrouter.ai/fusion • SWE-1.6: https://cognition.com/blog/swe-1-6 • SWE-bench: https://www.swebench.com/ • Tesla: https://www.tesla.com/ Where to find Russell: • LinkedIn: https://www.linkedin.com/in/russelljkaplan • Twitter/X: https://x.com/russelljkaplan Where to find Harrison: • LinkedIn: https://www.linkedin.com/in/harrison-chase-961287118/ • Twitter/X: https://x.com/hwchase17 Where to find LangChain: • Website: https://www.langchain.com/ • Docs: https://docs.langchain.com/ Send feedback or questions to maxagency@langchain.dev