OpenAI declared that the era of artificial general intelligence has begun, but many researchers remain unconvinced. The company's Astra model scored 99.9% through OpenAI's own software harness but only 62.7% on the standard ARC-AGI-3 benchmark. AI researcher François Chollet argues that the benchmark does not represent the finish line, since the real world involves far greater complexity and longer time horizons. There is still no universally accepted threshold for what counts as AGI.
President Donald Trump stated on Truth Social that he alone should oversee and regulate artificial intelligence, calling himself a strong and smart president. Vice President JD Vance reportedly shares this view. Several Democrats have proposed federal regulations for the AI industry in contrast. Tech CEOs are also urging the government to regulate AI due to potential harm, but Trump and Congress are delaying action. The AI industry has spent millions through political action committees to oppose federal rules and push for self-regulation.
Autodesk introduced new AI tools at AU 2026 for Fusion, Inventor, Vault, and PLM. The tools automate repetitive design and manufacturing tasks like assembly and timeline creation, aiming to reduce rework and help teams move faster from design to production. The company also unveiled a standalone, agent-first Autodesk Assistant planned for 2027, designed as an orchestration layer that routes tasks to the right AI tools. Autodesk also introduced Assistant Builder to create reusable and predictable agent workflows.
Goldman Sachs opened a new office in Bellevue, its first dedicated engineering space in the Pacific Northwest. The office will host more than 125 AI and cloud engineers. The company already has a Seattle office dating back to 2001 and employs over 12,000 engineers globally, about one-quarter of its total workforce. Meanwhile, a Florida couple learned that AI guidance can be risky without experience, ending up with extra wall holes on a home improvement project. General Manager Trent Stivers of Sunshine Ace Handyman Services says many callers start projects using online advice, and small jobs can grow into bigger issues when people miss important steps.
Idan Revivo, Head of Security Research at cyber firm Island, warns that AI lowers the cost of producing answers but raises the cost of trusting them. He believes security researchers must evolve into research engineers who can build analysis systems and verify their results. Island maintains a senior team of six researchers with complementary specialties to stay ahead of AI-accelerated threats. Separately, an Albright College professor says government oversight of AI makes sense, though some people calling for regulation may have other motives, and a writer criticizes recent writing conferences for leaving AI out of their programming.
Key Takeaways
- OpenAI says the AGI era has begun, but Astra model scored only 62.7% on the standard ARC-AGI-3 benchmark versus 99.9% in its own harness.
- Donald Trump declared on Truth Social that he alone should regulate AI, with JD Vance reportedly sharing the view.
- Multiple Democrats and tech CEOs are pushing for federal AI regulation while Trump and Congress delay action.
- The AI industry has spent millions through PACs to oppose federal rules and advocate self-regulation.
- Autodesk launched new AI tools at AU 2026 for Fusion, Inventor, Vault, and PLM to automate design and manufacturing tasks.
- Autodesk plans a standalone agent-first Autodesk Assistant for 2027, plus Assistant Builder for reusable agent workflows.
- Goldman Sachs opened a Bellevue office hosting more than 125 AI and cloud engineers, its first dedicated engineering space in the Pacific Northwest.
- Idan Revivo of Island warns AI lowers the cost of answers but raises the cost of trusting them, urging security researchers to become research engineers.
- A Florida couple's DIY home improvement experience shows AI guidance can lead to costly mistakes without hands-on experience.
- An Albright College professor supports AI regulation while noting some advocates may have other motives, and a writer urges conferences to include AI topics.
Autodesk brings new AI tools to Fusion and PLM at AU 2026
Autodesk announced new AI features at AU 2026 for Fusion, Inventor, Vault, and PLM. The tools automate repetitive design and manufacturing tasks like assembly and timeline creation. They aim to reduce rework and help teams move faster from design to production. The company stressed that AI grounded in project data can improve physical-world design.
Autodesk plans standalone AI Assistant for 2027
Autodesk unveiled a standalone, agent-first Autodesk Assistant planned for 2027. The tool uses project context like geometry and history to understand changes and their impacts. It is designed as an orchestration layer that routes tasks to the right AI tools. Autodesk also introduced Assistant Builder to create reusable and predictable agent workflows.
AI can help DIYers but cannot replace a handyman
A Florida couple used AI to guide home improvement projects but ran into problems. They ended up with extra wall holes and learned that AI guidance can be risky without experience. General Manager Trent Stivers of Sunshine Ace Handyman Services says many callers start projects using online advice. He explains that small jobs can grow into bigger issues when people miss important steps.
Albright professor discusses possible AI regulation
An Albright College professor says government oversight of artificial intelligence makes sense. The professor also notes that some people calling for regulation may have other motives. The article highlights the ongoing debate about how AI should be regulated.
Writing conferences need to include AI in their programs
A writer criticizes recent writing conferences for leaving AI out of their programming. The author says avoiding AI topics leads to confusion and rumor among attendees. They urge conferences to add AI panels, workshops, and clear policies. The article argues that open discussion helps writers deal with AI in publishing and creativity.
Island expert warns AI makes trust harder even as answers get cheaper
Idan Revivo, Head of Security Research at cyber firm Island, warns that AI lowers the cost of producing answers but raises the cost of trusting them. He believes security researchers must evolve into research engineers who can build analysis systems and verify their results. Revivo stresses that a security product is only as current as its understanding of attackers. Island maintains a senior team of six researchers with complementary specialties to stay ahead of AI-accelerated threats.
Trump says only he should regulate artificial intelligence
President Donald Trump declared on Truth Social that he alone should oversee and regulate artificial intelligence, calling himself a strong and smart president. Vice President JD Vance reportedly shares this view. Multiple Democrats have proposed federal regulations for the AI industry in contrast. The AI industry has spent millions through political action committees to oppose federal rules and push for self-regulation.
Goldman Sachs opens Bellevue hub for AI and cloud engineers
Goldman Sachs opened a new office in Bellevue, its first dedicated engineering space in the Pacific Northwest. The office will host more than 125 AI and cloud engineers. The company already has a Seattle office dating back to 2001. Goldman Sachs employs over 12,000 engineers globally, about one-quarter of its total workforce, and says the Bellevue location provides access to deep regional engineering talent.
Bellman Policy Optimization trains LLM reasoning without a critic
Bellman Policy Optimization reformulates LLM reasoning reinforcement learning as a critic-free trajectory objective. The method avoids estimating state values at intermediate states while sharing the same optimal solution as Policy Mirror Descent. This approach reduces variance, memory cost, and tuning complexity for sparse outcome-only rewards common in math, code, and tool-use tasks. Together with related research on granular credit assignment, it points toward simpler and more stable agent reinforcement learning.
OpenAI declares AGI era begun but researchers remain skeptical
OpenAI announced that the era of artificial general intelligence has begun, but many AI researchers are not convinced. There is no universally accepted threshold for what counts as AGI, and many dispute that current models have crossed it. OpenAI's Astra model scored 99.9% when tested through OpenAI's own software harness but only 62.7% using the standard ARC-AGI-3 benchmark setup. AI researcher François Chollet says the ARC-AGI-3 benchmark does not represent the finish line because the real world involves far greater complexity and longer time horizons.
Tech CEOs urge AI regulation as Trump and Congress delay action
Tech CEOs are urging the government to regulate artificial intelligence due to potential harm. However, Trump and Congress are not moving quickly to pass new rules. This political inaction creates a challenge for the tech industry. The warning comes from leaders in Washington. The technology continues to evolve rapidly.
Sources
- Autodesk advances AI for design and manufacturing at AU 2026
- Autodesk extends reach of AI with standalone Assistant
- Can artificial intelligence replace your handyman?
- Local professor shares thoughts on possible regulation of artificial intelligence
- Writing Conferences Shouldn’t Ignore AI in Their Programming
- Island: “AI made it cheap to produce an answer and expensive to trust one”
- 'High IQ' Trump Says Nobody But He Should Regulate Artificial Intelligence
- Goldman Sachs plants flag in Bellevue, opening hub for more than 125 AI and cloud engineers
- Bellman Policy Optimization Trains LLM Reasoning Without a Critic
- OpenAI says the AGI era has begun. AI researchers aren’t so sure
- Tech CEOs call for AI regulation. Trump and Congress are not rushing to act
Comments
Please log in to post a comment.