AUTOMATER NEWSROOM

What’s changing.
Why it matters.

New tools, useful workflows, and consequential AI developments.

Subscribe via RSS ↗

Latest briefing

Oct 28, 2026, 6:48 AM EDT
Discrete Visual Tokens Just Matched Continuous Ones on Multimodal Understanding — Rewriting a Core Assumption of Native Multimodal Models
A new arXiv paper shows quantization error is an unreliable proxy for semantic loss: with Factorized Vector Quantization, discrete visual token models become competitive with Qwen3-VL on OCR and understanding benchmarks.
Oct 28, 2026, 6:47 AM EDT
Cognition's FrontierCode 2.0 Grades Coding Agents on Vague Prompts. Opus 5.5 Tops It at 45.5%, and Max Effort Has a Cost
FrontierCode 2.0 uses deliberately underspecified prompts and adds three times as many thoroughness blockers. Here is what its scores, effort curves and failure examples mean for how you prompt, configure and review coding agents.
Oct 28, 2026, 5:58 AM EDT
Harmonic's Aristotle Now Formally Verifies AI-Generated C and Verilog, Not Just Math Proofs
Harmonic says Aristotle now formally verifies AI-generated C and Verilog, proving properties across every input and running checks in CI starting today.

Browse all →

Diagram showing a vague prompt leading to an agent patch, then blocking checks that give a score of zero on any failure or a weighted rubric score if all pass, above bars for FrontierCode 2.0 scores: Claude Opus 5.5 45.5%, GPT-6 Astra 42.8%, Grok 4.7 28.6%, Kimi K3 17.7%.

October 28, 2026

Cognition's FrontierCode 2.0 Grades Coding Agents on Vague Prompts. Opus 5.5 Tops It at 45.5%, and Max Effort Has a Cost

FrontierCode 2.0 uses deliberately underspecified prompts and adds three times as many thoroughness blockers. Here is what its scores, effort curves and failure examples mean for how you prompt, configure and review coding agents.

Our AI-assisted editorial method · Explore our in-depth Intel guides