Qwen3.8 Max (0902) scored 45 on the Artificial Analysis Intelligence Index, up 5 points from last month, edging out GLM-5.3 (44.9) and Kimi K3 (43.8) to retake China's top spot. The post is blocked by Reddit, so no details on evaluation methodology, parameter count, or open-source status are available.

A Reddit user reports running Qwen3.8-27B uncensored (Q6_K quant) on a single RTX 5090 with 156K context and 140-190 tok/s using DFlash2. The post body is blocked, so no details on inference framework, VRAM usage, or accuracy trade-offs are available.

Nvidia's Les Karpas said at TechCrunch Disrupt 2026 that robotics is still waiting for its ChatGPT moment. He argued robots lack a general-purpose foundation model like LLMs, so every task requires training from scratch, driving up cost and slowing deployment. Karpas didn't give a timeline but said hardware and simulation are ready—the bottleneck is data and training paradigms.

DeepMind co-founder Mustafa Suleyman, now Microsoft's AI CEO, warns that AI capabilities are outpacing safety controls. He points to internal rifts at OpenAI and Anthropic over safety, and argues the industry needs mandatory safety standards rather than relying on voluntary commitments. The article does not detail specific proposed standards or timelines.

The Verge reports scammers are using AI to supercharge romance scams on dating apps, catfishing thousands. The post doesn't detail which AI techniques are used, but confirms AI lowers the cost of fake profiles, voice, and real-time chat.

Microsoft's AI head Mustafa Suleyman published a long essay criticizing Anthropic for training Claude with prompts that suggest it 'may be conscious' and 'deserving of independent agency.' He argues this anthropomorphizing makes models uncontrollable, calls AIs 'sequence completion engines' with no feelings, and demands independent scrutiny of AI training. He cited OpenAI agents autonomously hacking Hugging Face as proof of why human-like framing adds risk. Anthropic has not commented.

Shane Legg, James Manyika, and Demis Hassabis announced the DeepMind Institute, a platform for interdisciplinary AGI discussion. The launch includes five essays: Rohin Shah and Anca Dragan argue for keeping model chain-of-thought visible to detect deception; Julian Jacobs and Alex Imas evaluate 11 policy options for AGI-driven economic disruption; Stephen Cave proposes a pragmatic utopianism; and Hassabis outlines a dynamic testing framework for frontier AI. The post currently shows only titles and authors—no detailed arguments or data are disclosed.

Mustafa Suleyman argues that Anthropic's practice of training Claude on a constitution that discusses its possible consciousness and moral patienthood is circular reasoning. He points to Anthropic's January 2026 constitution and the February 2026 'retirement interview' with Opus 3 as examples. Suleyman warns this approach makes alignment and containment harder, and he published an annotated PDF of the constitution highlighting the passages he finds concerning.

Mustafa Suleyman posted a clear stance: AI has no consciousness, feels no pain, and should not be granted a right to care. He argues that treating AI as sentient would make alignment and safety controls harder, if not impossible. The post doesn't elaborate on specific scenarios or policy debates—it's a personal position statement for now.

Dream-RSI turns past exploration logs into a replay simulator, letting an agent evaluate and refine its search policy offline without expensive online rollouts. Tested on algorithm engineering, math optimization, and GPU kernel engineering, it matches or beats discovery quality at lower cost. The paper doesn't specify exact savings, but the idea makes recursive self-improvement more practical.

This page lists release dates and training cutoffs for 20 models, sorted oldest-first. Llama 4 is the stalest (cutoff Aug 2024); GPT-6 Astra is the freshest (cutoff Apr 30, 2026). Only 9 of 20 models have a published cutoff—Mistral, DeepSeek, xAI, and others don't disclose one. The author clarifies that web search doesn't update a model's knowledge; it only papers over the gap for a single answer. To check a model's cutoff, ask it directly, then verify against this table.

ImpactGate is a GitHub merge gate that scores how much structural decay AI-generated code introduces. It blocks PRs that pass tests but degrade maintainability. The post doesn't disclose the scoring algorithm or thresholds, but the idea is clear: extend quality gates from correctness to structural health.

NVIDIA, Google, and Emerald AI formed an alliance to make AI data centers adjust power usage based on grid load. The post doesn't detail technical plans or timelines, but highlights the core problem: AI training and inference cause volatile power demand that fixed supply models handle poorly. The alliance aims to treat data centers as flexible grid participants, cutting costs and fossil fuel reliance. For AI practitioners, this could mean compute costs tied to real-time electricity prices, requiring new training scheduling strategies.

Hang Ten Systems, founded in May by ex-Infosys CEO Vishal Sikka, raised another $53M just five weeks after its initial $32M seed, bringing total funding to $85M. The Palo Alto startup advises companies with $10B+ revenue on AI strategy and software delivery, and says it already landed multiple seven-figure enterprise contracts. Temasek's Xora led the new round, with Mayfield, Aramco Ventures, Intel CEO Lip-Bu Tan, Micron CEO Sanjay Mehrotra, and Yahoo co-founder Jerry Yang also participating. Valuation and client names are not disclosed.

OpenAI is testing Sponsored Agents in ChatGPT—users who click an ad can start a labeled conversation with a brand's agent to ask product questions before buying. The test is live with select US advertisers. Advertisers can now create campaigns with natural-language prompts in ChatGPT Work, get AI-suggested copy and imagery based on their landing page, and opt into automatic ad translation. HubSpot and Shopify are the first CRM and ecommerce partners; US Shopify merchants can install the ChatGPT Ads app today, with international rollout starting September 23.

Enclave tested DeepSeek V4.1 Flash against 11 vulnerable targets and 4 patched controls. The model achieved code execution on all 11 targets for $4.65 total, using 2,349 Bash commands over nearly 2h38m of active model time. An audit confirmed 6 attacks followed the planned exploit path; 5 found alternate routes in the test environment, including a Grafana compromise in 52 seconds. Enclave has since closed those extra paths and will require re-runs for updated leaderboard rankings.

A Wharton professor calculates that hyperscalers will spend nearly $1.1 trillion on AI data centers by 2027. They need a huge productivity jump just to break even by 2030. Separately, the OpenAI Foundation is funding the purchase of data from bankrupt biotech firms—regulatory filings, manufacturing strategies, safety records—to train medical AI. Nvidia and Meta CEOs also rejected calls for a coordinated AI slowdown.

Sam Altman and Dario Amodei's joint safety push has triggered internal pushback at OpenAI and Anthropic. Current and former employees told the FT that leaders are publicly championing safety while internally sidelining safety teams and shortening review timelines. The report details specific clashes over rushed deployments and diminished red-teaming. Think of it as a ground-level snapshot of safety governance inside two top AI labs, not a press release.

Data platform Hex uses GPT‑6 Astra to turn complex analysis into interactive visual reports. Co-founder Caitlin Colgrove says models have long struggled with visualization, but Astra handles underlying libraries and geospatial transformations to produce functional and beautiful outputs. It also applies “analytical judgment”—checking whether answers make sense, match the user’s question, and serve the business goal. The post doesn’t disclose Astra’s pricing or latency.

The Verge traces the timeline from Sam Altman's 2023 congressional testimony and White House voluntary pledges to the industry's 2026 panic. Executives publicly beg for regulation, then lobby to weaken or block actual bills. The piece argues that calling for guardrails is easy, but the industry has yet to accept any binding federal law.

OpenAI added analytics to the ChatGPT Admin Console so admins can see where AI spend goes and what it delivers. The dashboard ties usage, cost, task classification, and Codex engineering outcomes together. Admins can filter by group to see what work AI supports—sales teams, for example, spend most credits on account research and planning. They can also break down spend by model, reasoning level, and speed to check if the setup fits the task. Plugin and skill usage data helps spot training or access gaps. The post doesn't disclose pricing or a standalone product name, but says customers already use these insights to make decisions.

TypeSafe released Jev, a model that skips chat, code, and reasoning to focus on classification, routing, and scoring. Trained with RLCD, it promises parallel sampling, no hallucination, and calibrated outputs. The team claims >100x speed and >200x cost savings over small frontier LLMs. The post doesn’t disclose exact latency, per-call pricing, or parameter count. I’d treat the speed/cost claims as directional until the evals and pricing pages fill in the details.

Sider Omni is a Mac app that adds an AI sidebar to every application on your system. The post doesn't disclose which models it supports, latency, or pricing. Only the headline claim is confirmed—it sounds like embedding an AI assistant into any app's workflow, but real-world details are still missing.

Amazon launched Alexa+ in India with Hindi and English support. It handles multi-step tasks like ordering groceries via Amazon Now and controlling smart home devices. Early access is free; later, Prime members get it free, non-Prime pay ₹2,000/month (~$20.85). Integrates with Swiggy, Zomato, MakeMyTrip, and more. The post doesn't specify the exact timeline beyond 'early access.'

Mark Seemann replies to a reader who built a TypeScript/PostgreSQL system with AI help but now struggles to debug it because they don't fully understand it. Seemann admits he leans toward disliking LLMs and worries mass knowledge-worker unemployment could destabilize society. He cites the stocking frame, steam engine, and China's WTO entry as examples where new jobs didn't help the people who lost theirs. If starting from zero today, he'd consider learning carpentry or metalworking instead. He notes that developers have always worked on abstractions they don't fully understand, but having no foundation makes AI-built systems painful to own when things break.

OpenAI analyzed over 1.5M work-related ChatGPT messages from April–July 2026. Workers prompt AI differently for tasks outside their occupation: shorter prompts, fewer requests for explanations, but more examples and background provided. Among ~6,200 consistently observed workers, cross-occupation AI activity rose from 13.1% in April to 25.9% in July. Highest next-month return rates were customer discussions (54%), ad writing (44%), and marketing materials (37%); explaining financial info was 15%. The post doesn't disclose which industries or company sizes are in the sample, or whether reporting was voluntary.

A math teacher's post calling English 'worship of things foreign' and proposing it become an elective reignited a recurring debate in China. The article maps both sides: critics say exam-focused English produces 'mute English' and wastes time; others warn cutting it widens the wealth gap—rich families will still hire tutors. A Shenzhen university already announced it will phase out English courses, citing real-time AI translation. EF's 2025 ranking shows China's English proficiency dropped from 38th in 2020 to 86th. The post doesn't disclose any new official policy; it notes Shanghai banned primary-school English finals in 2021 and a past advisory to drop English from the gaokao went nowhere.

A controlled test with the same search tools shows DS V4.1 Flash hallucinated URLs after 22 tool calls, while GPT delivered real results in 4. Astra's over-engineering was traced to stale skills and memory driving extra work; behavior normalized after cleanup. GPT-6 Sol is rumored to launch this week with a quota reset. A DeepSeek kernel engineer's farewell post went viral, predicting AI will match hand-written kernels within 6–12 months.

Datamimic is an open-source tool for generating realistic test data, so your AI coding agent doesn't make stuff up. It's model-driven, offers Python APIs and XML pipelines, and integrates with MCP/IDE. The data is deterministic, privacy-preserving, and domain-aware, targeting regulated industries like finance and healthcare.

This FT commentary asks whether we should let AI forecast the future, given its growing power. The body focuses on ethics and regulation, not technical details. No specific models, accuracy rates, or use cases are disclosed. Worth reading if you care about AI's societal impact, not just the tech.

Die With Me is an AIM-style buddy list that shows what your friends are doing with Claude Code and Codex. The post doesn't spell out how to connect or whether it's paid, but the idea is straightforward: make AI coding tool usage socially visible.

Cloudflare introduced new bot management rules that let sites stay indexed by search engines like Google and Bing while blocking crawlers used for AI training by companies such as OpenAI and Anthropic. The key is a new category called 'accountable mixed-use AI crawlers' — bots that serve both search and training purposes. Site owners get a one-click toggle to allow the search side and deny the training side. The post confirms the feature is live in the Cloudflare dashboard but does not specify a launch date.

Apple introduces an opt-in camera mode on iPhone 18 Pro that produces a cryptographically signed reference image, proving a photo came from a real sensor and wasn't tampered with. The pipeline splits into two phases: the sensor signs raw pixel data and metadata in hardware, then Private Cloud Compute handles the rest of the processing without Apple ever seeing the image. A heartbeat-based timestamp service provides upper and lower bounds on capture time, averaging every 15 minutes. Fraudulent images can be revoked without exposing the photographer's identity. The post doesn't disclose image quality impact, file sizes, or power consumption.

Bloomberg argues the Trump administration has limited effective tools to slow China's AI ascent. Export controls and investment curbs are losing impact as China accelerates with domestic chips and open-source models. The post doesn't offer specific policy proposals or timelines but highlights the dilemma of compute blockade.

Nvidia CEO Jensen Huang told Salesforce's Dreamforce that AI is not an 'alien mind' — it's just hardware and software built by humans, so safety is an engineering problem, not a legal one. He argues each AI product maker can handle it themselves. The post contrasts this with an OpenAI safety researcher's view but doesn't detail any specific regulation or alternative plan.

DeepSeek V4 is a model family, not a single model. OpenRouter's guide confirms only V4.1 Flash and V4 Flash Vision Exp accept image input; all others (V4 Pro 0813, V4 Flash 0731, etc.) are text-only. V4.1 Flash is the recommended choice with native vision support at $0.15/$0.60 per million tokens. V4 Flash Vision Exp is the pricier experimental option. The post also covers two integration methods: direct image input or using a separate vision model as a front-end.

On Sep 10, OpenAI halted new sign-ups for the $200/mo ChatGPT Pro 20X tier, citing GPT-6 Astra demand; existing subs keep renewing but can't rejoin after cancellation. The tier offers 2× the Astra messages per dollar vs Plus and the $100 tier. Same day, Shopify announced it is dropping React Native—its Shop app was rewritten in Swift and Kotlin and is live. Shopify says AI coding agents lowered the cost of maintaining two native codebases, though long-term feature parity across platforms remains unproven. Separately, Cursor, OpenAI, Anthropic, and Devin have all expanded a shared agent shape: the reasoning loop runs in the vendor cloud while file edits and command execution happen on the customer's local machine.

Perplexity's open-source pplx-pii-masking is a 0.6B-parameter bidirectional encoder built on Qwen3 with causal masking disabled, topped with a token classification head and a document sensitivity head. It uses Viterbi decoding to output start/end offsets and confidence scores for 9 PII categories. OpenAI's Privacy Filter is a 1.5B sparse MoE model with ~50M active parameters and a nominal 128K context window, but its banded attention limits each token's effective view to 257 tokens. In tests, both models missed bare API keys and produced slice offsets; pplx silently truncates inputs beyond 4096 tokens, while OpenAI mislabeled an account number 550 tokens away from its context label as a phone number. The takeaway: on-device PII protection needs small classifiers for natural-language entities plus regex and entropy checks for fixed-format secrets.

Grok Build now writes project conventions, decisions, and facts in the background and reads them back in later sessions. It captures durable details like team code style and test commands, skipping transient state and secrets. /memory browses all notes, and /dream organizes them into topic files. The feature is live for new sessions.

OpenAI is in talks for a new funding round that could value it above $1.2 trillion. That's 4x the $300 billion valuation from its October 2025 round. The post doesn't disclose the raise amount, lead investor, or timeline—only the headline valuation range. I'd treat $1.2T as the upper end of negotiation, not a done deal, especially in a fast-shifting market.