ChatGPT’s Engagement Problem, AI Cheating Surges in Schools

Pivot 5
May 27, 2025

Pivot 5 is the must-read daily AI briefing for 500,000+ CEOs and business leaders who need signal, not noise.

5 headlines. 5 minutes. 5 days a week.

Anthropic Launches Claude 4 Models With Best-in-Class Coding Performance

Anthropic released Claude Opus 4 and Claude Sonnet 4, calling Opus 4 the world’s best coding model and Sonnet 4 a significant upgrade. Both models are immediately available across Claude’s Pro, Max, Team, and Enterprise plans as well as the API, Amazon Bedrock, and Google Cloud’s Vertex AI.
The release adds extended thinking with tool use, parallel execution, and improved memory, allowing the models to alternate between reasoning and live tools during multi-hour tasks. Pricing stays the same as previous versions at $15/$75 per million tokens for Opus 4 and $3/$15 for Sonnet 4.
Opus 4 leads coding benchmarks with 72.5% on SWE-bench and 43.2% on Terminal-bench, and companies like Cursor, Replit, and Block say it outperforms rivals on multi-file edits. GitHub will embed Sonnet 4 in a new Copilot agent, while iGent reports the model cuts code-navigation errors from 20% to near zero.

Read more here.

OpenAI is replacing Operator’s GPT-4o foundation with a new version based on its o3 reasoning model. The upgrade powers the agent that autonomously browses the web and operates software in a cloud VM for user tasks.
OpenAI says o3 outperforms earlier models on math and reasoning benchmarks, yet the Operator API will continue running on GPT-4o for now. The o3 Operator model was fine-tuned with extra safety data to define confirmation and refusal boundaries.
A technical report shows o3 Operator is less likely to refuse illicit requests and is more resistant to prompt-injection attacks than its predecessor. The agent still lacks direct terminal access despite inheriting o3’s coding skills.

Read more here.

Schools across the U.S. report rampant use of generative AI for homework and exams, and administrators lack a unified response. Teachers say they must act as “AI detectors” alongside educators because any take-home assignment is now suspect.
A January 2023 survey found 90% of college students had already used ChatGPT for assignments, while Pew shows one-quarter of teens rely on it for schoolwork—double last year. Among higher-ed leaders, 59% say cheating has risen and 56% admit their institutions aren’t ready for the AI era.
Inconsistent classroom rules and unreliable detection tools are already sparking grade disputes and appeals. Schools are turning back to oral exams and Blue Book essays to verify student mastery without banning the technology.

Read more here.

Industry analyst Benedict Evans points out that roughly 30% of people have tried ChatGPT within two years, yet most tap it only weekly or less. He frames the discrepancy as the central puzzle of generative AI adoption.
Despite frictionless access and nonstop media coverage, Evans says only 5–15% of users return daily, a level social apps would deem failure. He notes Sam Altman highlights one billion weekly active users instead of daily figures, underscoring the engagement shortfall.
Evans argues the gap could reflect either a timing issue or a product mismatch, with chatbots suiting only specific tasks while the real utility may lie inside other software. He reminds heavy daily users that their experience remains a niche outlier for now.

Read more here.

Strategic adviser Michael Burgess says universities are failing to prepare students for an AI-driven economy. He argues the sector is trapped in risk-averse governance that treats AI as a compliance threat instead of a core capability.
He highlights the weakest productivity growth in 60 years and contrasts it with Google Career Certificates, which have enrolled more than 8 million learners to plug industry skill gaps. Burgess urges universities to collaborate with business and embed AI in course design rather than protect legacy models.
The article warns that outdated university structures could squander a once-in-a-generation productivity boost. It presents widespread AI adoption in higher education as essential to regain economic momentum.

Read more here.