Author
Adam Elkassas
Adam Elkassas writes at pre.dev about AI coding agents, agent harnesses and RL coding tasks, including the series on why most frontier coding tasks are useless for RL.
-
LabsMost “Frontier” Coding Tasks Are Useless for RL [Part 2]
How do long-horizon RL environments actually scale coding agents? In Part 2, pre.dev Labs breaks down custom GSPO, treating SFT as a rescue operation, and how catching bad signal early cut training costs by 5x–7x while quadrupling Terminal-Bench 3.0 scores.
-
LabsMost “Frontier” Coding Tasks Are Useless for RL [Part 1]
Do long-horizon RL tasks move the needle for coding agents? pre.dev Labs trained Qwen3.8-27B on ~100 frontier tasks, boosting its Terminal-Bench 3.0 score from 1/74 to 4/74. Here is how we built the environments and scaled agent performance.
-
AIThe unconventional path to stumbling on the best seed architecture for recursive self-improvement
The race for recursive self-improvement is on. How we combined long-horizon planning with coding agents to outpace the biggest frontier labs in AI.
-
AIFrontier labs won't build good harnesses. Their incentives won't let them.
Here's the problem nobody at Anthropic or OpenAI wants to say out loud: you can't build a token-efficient harness when your entire business model depends on burning tokens.Anthropic just committed…
-
AISelf-Driving Agentic Engineering: The End of Coding Agent Amnesia
In February 2026, Andrej Karpathy quietly retired the term he'd made famous. A year after coining "vibe coding" in a throwaway tweet that got 4.5 million views, he reframed the future as something…
-
BenchmarksThe 2026 Vibe Coding State of the Union: They All Suck
We took the most popular vibe coding tools at year-end 2025 and pitted them up against the new kid on the block: pre.dev. The results were unsettling (at least if you are an investor in Lovable and…
-
LaunchWe Built the Most Advanced AI Coding Agent—Here's Why It Actually Works
The vibe coding revolution is here. But most tools break the moment your project gets serious. We fixed that. Building software in 2025 looks nothing like it did two years ago. The rise of AI coding…
-
AISpecs as the compiler pass for AI code
You've probably been here: you give Claude or Cursor a detailed prompt to build a feature. It generates code. Looks good. You ship it. Then requirements change slightly, and suddenly the AI generates…
-
AIThe Enterprise AI Crisis: How Poor Project Scoping Is Killing Innovation
Enterprise AI projects face significant implementation challenges, with research indicating that 80% of AI initiatives fail to deliver their expected business value¹. Poor project scoping and scope…
-
AIThe $2M Mistake: Why 90% of Founders Get Project Scoping Wrong (And How AI Fixes It)
90% of startups fail¹, and a significant portion make the same costly mistake in project scoping during the early stages. They think they’re building an app. They’re actually building a company. And…
-
AIThe 2025 Vibe Coding Bubble
by Adam Elkassas, Founder pre.dev Why your credits keep disappearing and you still don’t have a working app. Vibe coding didn’t get you to the finish line, and you’re pissed. Now what? The 2025 Vibe…
Get new posts by email
New posts from pre.dev.