VC
Value Add VC
⚡HomePulse⚡Helpful Apps📝Blog
← Value Add PulseAI

New AI Optimization Framework Claims 2.5x More Output Than Claude Code and Codex on the Same Compute

Researchers introduced an optimization framework they say extracts roughly 2.5 times more useful output from coding agents like Claude Code and OpenAI's Codex while holding the compute budget fixed. If it holds up, it points to large efficiency gains sitting in orchestration rather than bigger models.

~2.5x output
Claimed Gain
Same compute budget
Constraint
Claude Code, Codex
Beats
Orchestration
Layer
TC
Trace Cohen
Early-stage VC & angel · Founder, New York Venture Partners
June 18, 2026
1 min read
KEY TAKEAWAYS FOR VCs & FOUNDERS
1

A 2.5x efficiency gain on fixed compute is a direct attack on the 'just scale up' thesis

2

It suggests orchestration and optimization, not raw model size, are where near-term gains hide

3

Cheaper effective inference reshapes the unit economics of every agent product

4

Framework-level wins are portable -- they ride on top of whichever model is best

TC
The VC Read · Trace's TakeTrace Cohen

If this replicates, it's a quiet shot at the entire 'just buy more GPUs' orthodoxy -- a 2.5x gain at the orchestration layer means a lot of headroom is sitting in how you run the model, not how big it is. That's great news for application founders and bad news for anyone whose whole pitch is access to scale. The catch is the word 'claims' -- benchmark-beating frameworks are a dime a dozen until someone independent reproduces them. But the direction is real: effective cost-per-task, not parameter count, is the number that decides what's profitable to automate.

🤖 AI Landscape →⚡ AI Chip Wars →

A newly published AI optimization framework reportedly delivers about 2.5 times more useful output than leading coding agents -- including Anthropic's Claude Code and OpenAI's Codex -- when each is held to the same compute budget. Rather than training a bigger model, the approach squeezes more out of existing ones by optimizing how the agent plans, allocates and executes its work.

The claim, if it survives independent scrutiny, lands on a sensitive nerve in the industry: the assumption that progress mostly comes from scaling parameters and buying more GPUs. A large, portable efficiency gain at the orchestration layer suggests a meaningful chunk of near-term improvement is available without touching the model at all.

“A large, portable efficiency gain at the orchestration layer suggests a meaningful chunk of near-term improvement is available without touching the model at all.”

The economic implications are the real story. Effective inference cost is the gating variable for agent products -- it determines what's profitable to automate. A framework that multiplies output per dollar reshapes those unit economics across the board, and because it rides on top of whatever model is strongest, it's the kind of advance that compounds with, rather than competes against, the frontier labs.

ShareXLinkedInEmail
More onAnthropic →OpenAI →

Originally reported by VentureBeat. Analysis and editorial commentary by Value Add Pulse.

← Back to Pulse

Markets Now

live
SPCX▲+2.52%
$224.10
CBRS▲+1.04%
$324.40
SPY▲+0.16%
5,931.80
QQQ▲+0.12%
19,972.10
NVDA▼-0.71%
$154.20
MSFT▲+0.25%
$477.30
GOOGL▲+1.22%
$207.90
META▲+0.25%
$651.40

Read Next

AINobel laureate hire

Nobel Laureate John Jumper Leaves Google DeepMind for Anthropic, Deepening Google's AI Talent Drain

John Jumper -- the 2024 Nobel laureate who co-created AlphaFold -- is leaving Google DeepMind after nearly nine years to join rival Anthropic, just days after Gemini co-lead Noam Shazeer departed for OpenAI. Back-to-back marquee exits have turned the AI race into an outright talent war, and DeepMind is the lab bleeding.

AI

Anthropic's Claude Code Artifacts Update Brings Live, Shared Dashboards to Enterprises

Anthropic updated Claude Code Artifacts to support live, shared dashboards and interactive workspaces, turning what was a single-user coding tool into a collaborative surface teams can build on together. It's a push to make Claude Code the place enterprises do work, not just generate code.

AI

Signal's Meredith Whittaker Warns AI Chatbots 'Are Not Your Friends'

Signal president Meredith Whittaker is pushing back hard on the rise of companion-style AI, warning that chatbots designed to feel like friends are surveillance-driven products optimized for engagement, not your interests. Her argument lands as companion AI goes mainstream and regulators start to circle.

@Trace_Cohen·t@nyvp.com