The trick: Rented Halo

Claude hit 14 of 15 protein targets, and outside labs confirmed it.

Then read the method: Claude drove the specialist design tools the field already ships, and a binder is the first step of a drug, not the drug.

Issue 1319 August 20266 receipts4 min

Anthropic reports that Claude, using Opus 4.8 and a Mythos Preview, designed protein binders against 14 of 15 targets, with 22 to 35 percent of individual designs binding successfully against a typical 10 to 15 percent, all physically produced and tested by two outside labs, Adaptyv Bio and Twist Bioscience.

Before you read on. Your call?

unusually for this beat, the numbers hold. The designs were made and measured by independent evaluators, and the baseline is sourced. What the headline hides is the method. Claude did not invent a protein-design engine. It orchestrated the structure, sequence, and co-folding models the field already uses, asked to produce 30 candidates per target. Its per-design hit rate of 22 to 35 percent, averaged across all 15 targets, beats the loose field average and sits below the specialist best case of 70 to 80 percent, though that figure comes from just two curated targets with pre-filtered designs, so it is not a clean head to head. And Anthropic states plainly that a minibinder is not a standard drug and that a high-affinity binder is only the first step. This is a strong agent-orchestration result being read as a molecular-biology breakthrough.

14/15TARGETS CLAUDE GOT A BINDER AGAINST

0NEW DESIGN MODELS CLAUDE BUILT

22-35%CLAUDE PER-DESIGN RATE ACROSS 15 TARGETS

STEP 1WHERE A BINDER SITS ON THE ROAD TO A DRUG

There’s more to this story.

Membership opens the full investigation, the strongest counterargument and what to do with what you’ve learned.

Start your free month →

First membership: 30 days free, then A$89 a year. One introductory trial per customer. Card required; renews annually until cancelled. Cancel before the trial ends to avoid the first charge. Already a member? Sign in

The trick has a name

We call it Rented Halo: the achievement is real, the drama around it is borrowed. You'll see it again. Learn to spot it →

Say this in tomorrow's meeting“Claude got binders against 14 of 15 targets and outside labs confirmed the binding, which is real. But Claude orchestrated the specialist tools the field already had, its 22 to 35 percent rate is an average across 15 targets while the specialist 70 to 80 percent is a best case on two curated ones, and Anthropic itself calls a binder just the first step toward a drug.”

Receipts

Supports anthropic.com: of the 15 targets we designed against, Claude successfully designed binders against 14 of them
Supports anthropic.com: Between 22% and 35% of its individual designs bound successfully, depending on the setup, compared to the 10-15% that is typical in protein design campaigns today.
Context anthropic.com: Claude did this by operating publicly available specialist protein design and co-folding models that the field already uses.
Context anthropic.com: Protein minibinders are not a standard therapeutic modality for drugs and even for
Context arxiv.org: achieving hit rates of 70%-80% and picomolar-level binding affinities
Context pmc.ncbi.nlm.nih.gov: RFdiffusion has demonstrated outstanding performance in designing picomolar-affinity helical peptide minibinders, symmetric structural assemblies, and various minibinders targeting antigens such as GAPD and TNFR

Open the Receipts Pack → What each source proves, every figure traced, and what would change our verdict.