Researcher Derik Kauffman found an iOS list that blocks about 100,000 words and phrases from swipe typing and predictive suggestions.
Browsing Category
Technology and Science
437 posts
How T3 Code Brings Claude Code And Codex Subscriptions Into One Workspace
The free, open-source T3 Code app brings multiple coding agents into one interface while users keep their existing provider subscriptions.
How AI Agents Remember: What Anthropic’s Memory Research Shows, And What We Learned Running Agents With Memory
Anthropic’s November 6 research post describes how Claude agents save, organize and correct memories between sessions, with results from company-run tests.
Terminal-RSI, Checked: Did An AI Agent Really Improve Itself With No Humans Involved?
Terminal-RSI reports benchmark gains from a human-designed training pipeline. The paper does not show an AI improving itself without human involvement.
What Did Chainalysis Mean By “Non-Sensitive Data” In Its Breach Updates?
Chainalysis first described accessed data as non-sensitive, then disclosed password hashes and live tokens for all users. Several details remain unknown.
Is AI Now Speeding Up Its Own Development? What METR Found Inside Anthropic, And What Stayed Redacted
METR estimates Anthropic’s model progress accelerated 1.9x, with AI use likely a major factor. Much of the evidence remains redacted.
Most Companies Use AI. Only About 6 Percent Profit From It
Surveys find broad AI adoption but limited profit impact: 6% of McKinsey respondents attribute at least 5% of EBIT to AI.
47 Local Ballot Votes, 46 No Votes: What The Results Mean
Unofficial results show restrictions leading in 46 of 47 local data center measures across 10 states, though the measures vary widely.
How We Checked OpenAI’s Claim That ChatGPT Handles A Fifth Of Google’s Searches
OpenAI says ChatGPT handles about one-fifth of Google’s queries. Independent estimates and traffic panels show different measures and much smaller website traffic.
We Gave 42 AI Setups The Same 15 Legal Questions. The Model Built For Law Scored 8 Out Of 100
A 15-question Lexassay test scores 42 AI setups on legal tasks. Claude Opus 5.5 leads; a model built for law scores 8 out of 100.