Blog - Adam Holter

May 19, 2026

Google I/O 2026 Leaks: Gemini 3.5, Omni, and What to Expect

“Google is a leaky bucket.” Just like their Pixel phones always leak, there have been tons of leaks leading up to Google I/O about what we are going to get there. Some of them were intentional by the team, others less so. The pattern fits everything from UI strings appearing in production then vanishing to […]

May 13, 2026

Google Caught an AI-Generated Zero-Day Exploit

Pay attention to Google’s latest Threat Intelligence Group report. It documents what appears to be the first case...

ai-news ai-trends

May 5, 2026

You Shouldn't Use GPT-5.5 Instant

You shouldn’t use GPT-5.5 Instant. The reason is straightforward. You should not use any instant models if you...

ai-models chatgpt

May 5, 2026

SubQ and Long Context Efficiency

SubQ claims the first fully subquadratic LLM tuned for 12 million token reasoning without hybrids or quality drop....

ai-models ai-research

May 5, 2026

I Should Have Applied to the GPT-5.5 Party Anyway

I should have applied to the GPT-5.5 party anyway. The RSVP form went live and I passed because...

ai-news cost-efficiency

May 4, 2026

Blueprint-Bench 2 Reveals the First Signs of 3D Spatial Intelligence

Andon Labs released Blueprint-Bench 2. The results show the first measurable signs of 3D spatial intelligence in frontier...

ai-research benchmark

May 1, 2026

This GPT Image 2 Prompt for Deliberately Bad Drawings

This GPT Image 2 prompt is going insanely viral right now. The full text tells the model to...

ai-tools generative-ai

May 1, 2026

OpenAI Traced Their Goblin Problem to One Personality Reward

OpenAI traced their models increasing use of goblin and gremlin metaphors back to rewards given during training for...

May 1, 2026

Here’s Why I’m Not Switching to the $100 ChatGPT Plan

I pay for both the twenty dollar OpenAI plan and the one hundred dollar Claude plan right now....

ai-tools chatgpt

Apr 28, 2026

GPT-5.5 Had to Ban Goblins Twice

OpenAI told GPT-5.5 twice not to talk about goblins. The Codex models.json file contains the exact same instruction...

ai-models ai-trends

Apr 18, 2026

Claude Design Review: Build Prototypes and Slides with Claude Opus 4.7

Anthropic launched Claude Design on April 17, 2026. The tool lets users create prototypes, slides, one-pagers and branded...

ai-tools anthropic

Apr 18, 2026

Google I/O 2026 Leaks: Gemini 3.5, Omni, and What to Expect

Google Caught an AI-Generated Zero-Day Exploit

You Shouldn't Use GPT-5.5 Instant

SubQ and Long Context Efficiency

I Should Have Applied to the GPT-5.5 Party Anyway

Blueprint-Bench 2 Reveals the First Signs of 3D Spatial Intelligence

This GPT Image 2 Prompt for Deliberately Bad Drawings

OpenAI Traced Their Goblin Problem to One Personality Reward

Here’s Why I’m Not Switching to the $100 ChatGPT Plan

GPT-5.5 Had to Ban Goblins Twice

Claude Design Review: Build Prototypes and Slides with Claude Opus 4.7

Opus 4.7 Correctly Identifies Kelsey Piper From Unpublished Heist Novel Text

GPT-5.4 Beats Claude 4.6 Opus in 5D Chess

CRUX Tests Open-World AI Evaluations With Live iOS App Publication

I'm a 1% Addict: Why Small LLM Benchmark Gains Matter

OpenAI Codex Super App Launches Same Day as Claude Opus 4.7

Claude Opus 4.7 Delivers Autonomy Gains on Hard Coding Tasks

Claude Code Desktop Redesign Adds Multiple Sessions Side By Side

OpenAI Report: Why AI Matters for Life Sciences

GPT Image 2: Mobile App Strings and Renewed LM Arena Tests Point to Near-Term Release

Claude Code Routines: What the Research Preview Actually Delivers for Cloud Automation

Gemini Robotics-ER 1.6 Delivers Targeted Gains in Robot Vision and Safety

Claude Mythos Preview: First AI to Complete AISI 32-Step Cyber Range End-to-End

Model Degradation: Intentional Sabotage or Accidental Slips?

AI Backlash Hits Home: Sam Altman's Residence Faces Multiple Attacks