TypeSafe AI emerged from stealth with Jev, a model designed to return structured, type-safe decisions rather than conversational text. Built by OpenAI veteran Diogo Almeida over two years, Jev uses a training approach called Reinforcement Learning for Calibrated Decisions. The seed round raised about $40 million led by DCVC, and the model prices input at $0.042 per million tokens with output free. Early tests show response times between 70 and 500 milliseconds and speedups up to 193.6 times on certain workflows.
OpenAI is expanding ChatGPT Ads with AI-powered features including Sponsored Agents. These agents let users interact with business-sponsored entities after clicking an ad, running separately from standard ChatGPT responses. HubSpot and Shopify are the first CRM and ecommerce partners integrated with the platform. The Sponsored Agents test is currently limited to the US and select partners, and AI-generated suggestions still require human approval.
Snap is partnering with Salesforce, Nvidia, and Amazon Web Services to bring workplace tools to its Specs augmented-reality glasses, priced at $2,195. The partnerships aim to attract business customers in a market currently dominated by Meta. Meanwhile, 10x Genomics teamed up with Lunit to connect AI pathology engines to its cancer biomarker discovery platform, as the company's stock has surged over 347% year-to-date.
On the policy and security front, Exabeam found that 48% of security leaders now view AI agents as the biggest insider threat, ahead of external hackers. Financial advisers are also adopting AI quickly, with 85% using it and 55% saving at least four hours weekly. Reid Hoffman criticized current AI messaging at POLITICO's Decoded Summit, while House Energy and Commerce Chair Brett Guthrie warned against strict European-style regulations, saying the US cannot afford to fall behind China in AI development.
Key Takeaways
- <ul><li>TypeSafe AI launched Jev, a model that outputs structured decisions instead of chat, with a $40 million seed round led by DCVC</li><li>OpenAI introduced Sponsored Agents for ChatGPT Ads, partnering with HubSpot and Shopify, currently in US-only testing</li><li>Snap partnered with Salesforce, Nvidia, and AWS to add workplace tools to its $2,195 Specs AR glasses, competing with Meta</li><li>Exabeam found 48% of security leaders consider AI agents the top insider threat, above external hackers</li><li>85% of financial advisers now use AI, with 55% saving at least four hours per week, according to AssetMark</li><li>10x Genomics partnered with Lunit on AI pathology tools as its stock gained over 347% year-to-date</li><li>Reid Hoffman criticized AI messaging at POLITICO's Decoded Summit, saying candidates explaining economic benefits could gain voter support</li><li>House Energy and Commerce Chair Brett Guthrie warned against European-style AI regulations at the POLITICO Decoded Summit</li><li>Anthropic faces disclosure scrutiny over a potential greater than 10 percent chance of human extinction from its product</li><li>27% of companies still lack strong monitoring capabilities for AI despite 60% using dedicated AI security tools</li></ul>
TypeSafe AI launches Jev model that returns decisions not chat
TypeSafe AI came out of stealth on September 15, 2026 with Jev, a System One model that returns typed decisions instead of chat. The seed round was about $40 million led by DCVC. Jev is priced at $0.042 per million input tokens with output tokens free. The model uses a training method called Reinforcement Learning for Calibrated Decisions and targets use in software workflows.
ChatGPT co-inventor launches Jev for structured decisions in code
Jev was built over two years in stealth by OpenAI veteran Diogo Almeida. It takes unstructured state as input and outputs type-safe structured values in a single parallel query. The model uses Reinforcement Learning for Calibrated Decisions and a hardware-aware parallel sampler. Tests show response times of 70 to 500 milliseconds and speeds up to 193.6 times faster on certain workflows.
OpenAI introduces AI-powered tools for ChatGPT Ads
On September 16, 2026 OpenAI introduced new AI-powered ad experiences for ChatGPT Ads. These include Sponsored Agents that let users talk with business-sponsored agents after clicking an ad. Advertisers can now use natural-language prompts in ChatGPT Work to create and manage campaigns. HubSpot and Shopify are the first CRM and ecommerce partners integrated with the ad platform.
OpenAI tests Sponsored Agents that respond inside ChatGPT
On September 16, OpenAI tested Sponsored Agents that run separately from ChatGPT answers. Shoppers can ask sponsored agents follow-up questions about products seen in ads. HubSpot merchants can connect Sponsored Agents to campaigns. Sponsored Agents are currently a US-only test with select partners, and AI-generated suggestions still need human approval.
FT Alphaville examines Anthropic's existential risk factor
FT Alphaville published a piece about Anthropic's risk factors. The article highlights the possibility that Anthropic's product has a greater than 10 percent chance of causing human extinction. It frames this as a key disclosure issue for the company's future financial filings.
10x Genomics partners with Lunit on AI pathology tools
10x Genomics announced a partnership with Lunit to connect AI pathology engines to its cancer biomarker discovery tools. The company's stock has risen sharply, gaining 347.59% year-to-date and 466.57% over one year. Shares closed at $74.39, well above a fair value estimate of $49.54. New products like Flex v2 and Visium HD are expected to expand the company's market reach.
Snap partners with Nvidia and Salesforce for AR work glasses
Snap announced partnerships with Salesforce, Nvidia, and Amazon Web Services to add workplace tools to its Specs augmented-reality glasses. The Specs glasses, priced at $2,195, were first shown in June. The move targets business customers in a market currently led by Meta. The company hopes these partnerships will help it compete in the crowded AR space.
AI agents seen as top insider security risk by Exabeam
Exabeam research found that security leaders now view AI agents as the biggest insider threat to organizations. A survey of 600 decision-makers across seven countries found that 48% ranked AI agents above external hackers. AI agents are defined as goal-driven autonomous systems, not chatbots. While 60% of companies use dedicated AI security tools, 27% still lack strong monitoring capabilities.
AI helps financial advisers save time but brings errors
A report by AssetMark found that 85% of financial advisers now use AI, with 55% saving at least four hours per week. Advisers most often use AI for drafting emails, research, and taking meeting notes. However, 37% cite home office hesitance and compliance as major barriers to adoption. Experts say firms need to move beyond task assistance to full automation to gain real benefits.
Reid Hoffman says AI messaging has failed with Americans
LinkedIn co-founder Reid Hoffman criticized how AI has been communicated to the public. He spoke at POLITICO's Decoded Summit in Washington and said candidates who explain AI's economic benefits could find voter support. Hoffman said current messaging around data centers is not working. He added that he does not require candidates to be pro-AI to earn his support.
AI Skeptics Doubt Slowdown Plan
The article discusses skepticism from AI critics about proposed slowdowns in artificial intelligence development. It references a piece by Troy Wolverton at The Examiner. The article invites tips about tech, startups, and the venture industry.
Guthrie Warns Against Surrendering AI Race With China
House Energy and Commerce Chair Brett Guthrie warned against strict European-style regulations on artificial intelligence. He spoke at the POLITICO Decoded Summit in Washington on Wednesday. Guthrie said such rules could limit competition and hurt the United States in its global AI race with China. He stressed that the country absolutely cannot surrender its AI development progress.
Sources
- TypeSafe Jev: the frontier model built to kill chat
- ChatGPT pioneer launches Jev model for programmatic logic
- Reimagining advertising with AI
- OpenAI lets ads talk back inside ChatGPT
- Anthropic’s mother of all risk factors
- Is 10x Genomics (TXG) Fully Priced On Its New Lunit AI Pathology Partnership?
- Snap targets enterprises with Salesforce, Nvidia AI tools for augmented-reality glasses
- AI agents now seen as biggest insider risk, Exabeam
- AI Brings Gains for Time Management, but More Mistakes, Too
- 'AI has been marketed horribly to the American people,' LinkedIn co-founder says
- AI skeptics dubious of proposed slowdown
- 'We absolutely can’t surrender,' Guthrie says of AI race with China
Comments
Please log in to post a comment.