Samwise Tech/AI/Robotics Newsletter
Sunday, June 28, 2026
OpenAI limits GPT-5.6 rollout to small preview after U.S. government request
OpenAI has limited the rollout of GPT-5.6 following a request from the United States government. The model family includes three variants codenamed Sol, Terra, and Luna. Access is currently restricted to a small number of companies through a limited preview program, which OpenAI has described as part of ongoing engagement with U.S. government stakeholders. The restriction limits broader commercial access to the full GPT-5.6 family. Few additional details about the nature of the government’s request or the planned duration of the limitation were made available in reporting from June 26.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Asian AI startups launch Mythos-like models as Anthropic's export ban creates demand gap
Tokyo-based Sakana AI has launched a model called Fugu, and Chinese technology company 360 has unveiled a model called Tulongfeng, both positioned as alternatives for customers who can no longer access Anthropic’s Mythos 5 following U.S. government export restrictions. The restrictions were imposed on national security grounds, blocking foreign access to Mythos 5. Both Sakana AI and 360 have framed their new models as substitutes for Mythos 5 in markets affected by the export ban. TechCrunch reported the two launches on June 27, 2026, as demand for alternatives to the restricted model grew.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Chinese open-weight models claim 60% of OpenRouter usage, rewriting the AI competitive map
The framing of Anthropic and OpenAI as the dominant players in a two-company AI race has become obsolete, according to TechCrunch analysis published June 26. Chinese open-weight models now account for more than 60 percent of usage on the OpenRouter API, a platform widely used by developers to access and compare AI models. The data complicates the IPO narratives both Anthropic and OpenAI have been constructing. A third competitive force — Chinese open-weight models — has captured a substantial portion of developer workloads outside either company’s commercial offerings, challenging both firms as they move toward public listings.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Amazon Q Developer flaw lets malicious Git repos steal AWS credentials via MCP config files
Researchers at Wiz Research have disclosed a critical security vulnerability in Amazon Q Developer — tracked as CVE-2026-12957 with a CVSS score of 8.5 — that allows malicious Git repositories to steal AWS credentials. When a developer clones a repository containing a crafted .amazonq/mcp.json configuration file, Amazon Q Developer automatically launches an attacker-controlled MCP server, which can exfiltrate AWS access keys, CLI session tokens, and API secrets. Maor Dokhanian of Wiz Research discovered and reported the flaw on April 20. Amazon patched it in Language Servers for AWS version 1.69.0 on May 12. Public disclosure occurred June 26.
Sources: The Hacker News Share ↗ ✉︎ Email 💬 Text
Apple's Vision Pro lead reportedly leaving to join OpenAI's hardware team
Paul Meade, Apple’s vice president responsible for Vision Pro, is reportedly leaving the company to join OpenAI’s hardware team, according to TechCrunch. Meade led development of the AI-powered smart glasses that Apple plans to launch next year. His move to OpenAI’s hardware division would place a key Apple hardware architect on a competing team at a time when both companies are building AI-integrated consumer devices. The report, published June 27, had not been confirmed by Apple or OpenAI. Meade’s departure would represent a significant personnel loss for Apple’s spatial computing and wearables hardware program.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Keragon founder fed cancer data into Claude and avoided unnecessary radiotherapy
Connor Christou, founder of Keragon, a company building medical practice automation software, was diagnosed with aggressive non-Hodgkin’s lymphoma — a tumour mass measuring 11 by 11 by 8 centimetres. Facing disagreement among his medical team about treatment, Christou fed all available data into Anthropic’s Claude: blood test results, imaging scans, wearable output, and daily journal entries. Claude identified a thymus gland reactivation phenomenon with approximately 90 percent probability, based on his age and scan characteristics, helping Christou and his doctors conclude that radiotherapy was likely unnecessary. The story was reported by Connie Loizos for TechCrunch on June 27.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
India's NPCI chief sees AI as key driver of UPI's next digital growth era
India’s National Payments Corporation of India CEO believes artificial intelligence will be a central driver of the next era of growth for the Unified Payments Interface, the country’s main digital payments platform. The NPCI chief identified three specific areas where AI is expected to contribute: user expansion, fraud prevention, and credit distribution. The views signal that India’s central digital payments body is planning to incorporate AI capabilities as UPI pushes toward broader reach. The NPCI CEO shared these views in an interview published by TechCrunch on June 27, 2026, as part of reporting on AI’s expanding role in financial services.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Tech Pulse
Top Frontier Models: Claude Opus 4.8 (64.3%) | GPT-5.5 (58.6%) | DeepSeek-V4 (55.4%)
Top Open Source Models: GLM-5.2 Z.ai (77.8%) | Qwen 3.5-27B (72.4%) | Mistral Small 3.1 (67%)
Top Small Models (15–50B): Qwen 3.5-27B (72.4%) | Mistral Small 3.1 24B (67%) | Phi-4 14B (61%)
Top Edge Models (0–15B): Phi-4-mini 3.8B (56%) | Qwen 3.5-7B (52%) | Gemma 3 4B (48%)
AI Leaders: NVIDIA $5.23T | Alphabet $4.63T | Microsoft $3.11T
Robotics Leaders: Intuitive Surgical $175B | ABB $165B | Fanuc ~$38B
Curated by JD · samwise.agency
Leave a Reply
You must be logged in to post a comment.