Every AI story that matters — and the intelligence behind it.
TLDRocket reads all relevant sources, removes duplicate coverage, and publishes a short neutral
summary of every story, linking back to the original. Free, no spam, unsubscribe anytime.
Google released MedGemma 1.5 4B, an updated medical AI model that now handles high-dimensional medical imaging like CT and MRI scans alongside text and 2D images, and launched MedASR, a speech-to-text model fine-tuned for medical dictation. MedGemma 1.5 improved accuracy by 3% on CT classification and 14% on MRI classification over its predecessor, while MedASR achieved 58% fewer errors than Whisper on chest X-ray dictations. Google is hosting a $100,000 MedGemma Impact Challenge hackathon and making both models freely available on Hugging Face and Google Cloud for developers to build healthcare applications.
Google Quantum AI demonstrated dynamic surface codes for quantum error correction on their Willow processor, where error-correcting circuits alternate between different configurations rather than using fixed static circuits. The research showed three variants—hexagonal, walking, and iSWAP codes—that each address specific hardware challenges, with the hexagonal code requiring only three couplers per qubit instead of four and achieving a 2.15× improvement in logical error rate as code distance increased from 3 to 5. Dynamic circuits enable greater flexibility in gate selection, connectivity patterns, and error suppression while avoiding qubit dropouts, moving toward the goal of logical qubits with error rates below one error per million correction cycles.
Google released Veo 3.1 Ingredients to Video, an update to its video generation tool that creates clips from reference images with improved character and background consistency. The tool now supports native 9:16 vertical video format and upscaling to 1080p and 4K resolution, with availability across YouTube Shorts, the Gemini app, Flow, Google Vids, and Vertex AI. Videos generated by the system are embedded with SynthID watermarks and can be verified within the Gemini app to identify AI-generated content.
Zenken deployed ChatGPT Enterprise across its organization to improve sales operations and customer engagement. The company reduced proposal preparation time and increased proposal success rates, though specific metrics were not disclosed. Sales staff can now handle more personalized customer work with the same lean team size.
Salesforce launched a rebuilt Slackbot AI agent that transforms the workplace chat tool from a basic notification system into a full AI assistant capable of searching enterprise data, drafting documents, and taking autonomous action. Internal testing with 80,000 Salesforce employees showed two-thirds tried the new Slackbot, with 96% satisfaction among users and employees reporting savings of two to 20 hours per week. The release positions Slack as Salesforce's centerpiece for competing with Microsoft's Copilot and Google's Gemini in enterprise AI, with plans to support multiple AI models and coordinate with third-party agents.
Sequoia led a pre-seed and seed investment in Sandstone after meeting founder Nick Fleisher and co-founder Jarryd Strydom, framing Sandstone as an AI-native workflow engine for in-house legal teams. Sequoia met Fleisher and then led Sandstone’s pre-seed two months later, with the MVP reportedly built in weeks. The funding and accelerated build led Sandstone to scale its product rollout with early customers and continue toward larger expansion.
WithCoverage partnered with industry risk specialists to use an AI audit engine that analyzes a business’s policies and operations to produce line-by-line findings on risks and insurance savings. It says customers can save up to 30% on their total insurance budget, potentially reaching seven figures. The market process shifts from email/PDF workflows and opaque commissions to a digital platform with transparent fees and more personalized risk assessments, starting from a customer’s first call.
Cursor partnered with Together AI to optimize real-time code completion inference using NVIDIA Blackwell hardware and advanced quantization techniques. The partnership productionized B200/GB200 chips with FP4 quantization and TensorRT optimizations to achieve low-latency inference at scale. This infrastructure enables Cursor's in-editor AI agents to maintain fast response times while processing requests reliably.
Every AI story that matters,
in your inbox by 8am.
TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the
day in two minutes. Follow companies and topics for alerts, or get the
briefing in Slack. Free, no spam, unsubscribe anytime.
Reading TLDRocket needs no cookies, and the readership counts we rely on come from
our own cookieless analytics. Google Analytics is the exception: it sets cookies and
reports to Google, so it stays switched off until you allow it. You can change your
mind any time from “Cookie settings” in the footer.
Strictly necessary
Session security and form protection (tldrocket-session,
XSRF-TOKEN, 2 hours). The site cannot work without them,
so they need no consent.
Always on
Google Analytics 4 (_ga,
_ga_<id>, up to 2 years). Measures which
stories and sections readers use. Google acts as a third-party processor and may
store the data outside the EU. No advertising, no profiling, no data sold.
