November ended. Thanksgiving (within the US), turkey, and a prepare of mannequin bulletins. The bulletins have been thrilling: Google’s Gemini 3 places it within the lead amongst giant language fashions, no less than in the intervening time. Nano Banana Professional is a spectacularly good text-to-image mannequin. OpenAI has launched its heavy hitters, GPT-5.1-Codex-Max and GPT-5.1 Professional. And the Allen Institute launched its newest open supply mannequin, Olmo 3, the main open supply mannequin from the US.
Since Developments avoids deal-making (ought to we?), we’ve additionally averted the angst round an AI bubble and its implosion. Proper now, it’s protected to say that the bubble is shaped of cash that hasn’t but been invested, not to mention spent. If it’s a bubble, it’s sooner or later. Do guarantees and needs make a bubble? Does a bubble made from guarantees and needs pop with a bang or a pffft?
AI
- Now that Google and OpenAI have laid down their playing cards, Anthropic has launched its newest heavyweight mannequin: Opus 4.5. They’ve additionally dropped the value considerably.
- The Allen Institute has launched its newest open supply mannequin, Olmo 3. The institute’s opened up the entire improvement course of to permit different groups to know its work.
- To not be outdone, Google has launched Nano Banana Professional (aka Gemini 3 Professional Picture), its state-of-the-art picture technology mannequin. Nano Banana’s largest characteristic is the flexibility to edit pictures to alter the looks of things with out redrawing them from scratch. And in keeping with Simon WIllison, it watermarks the components of a picture it generates with SynthID.
- OpenAI has launched two extra parts of GPT-5.1, GPT-5.1-Codex-Max (API) and GPT-5.1 Professional (ChatGPT). This launch brings the corporate’s strongest fashions for generative work into view.
- A bunch of quantum physicists declare to have lowered the dimensions of the DeepSeek mannequin by half, and to have eliminated Chinese language censorship. The mannequin can now let you know what occurred in Tiananmen Sq., clarify what Pooh seemed like, and reply different forbidden questions.
- The discharge prepare for Gemini 3 has begun, and the commentariat rapidly topped it king of the LLMs. It contains the flexibility to spin up an online interface so customers can provide it extra details about their questions, and to generate diagrams together with textual content output.
- As a part of the Gemini 3 launch, Google has additionally introduced a brand new agentic IDE known as Antigravity.
- Google has launched a brand new climate forecasting mannequin, WeatherNext 2, that may forecast with resolutions as much as 1 hour. The information is on the market by means of Earth Engine and BigQuery, for many who want to do their very own forecasting. There’s additionally an early entry program on Vertex AI.
- Grok 4.1 has been launched, with experiences that it’s presently the most effective mannequin at generative prose, together with artistic writing. Be that as it might, we don’t see why anybody would use an AI that has been skilled to replicate Elon Musk’s ideas and values. If AI has taught us one factor, it’s that we have to suppose for ourselves.
- AI calls for the creation of latest information facilities and new vitality sources. States need to guarantee that these energy crops are constructed, and in-built ways in which don’t cross prices on to customers.
- Grokipedia makes use of questionable sources. Is anybody shocked? How else would you prepare an AI on the newest conspiracy theories?
- AMD GPUs are aggressive, however they’re hampered as a result of there are few libraries for low-level operations. To unravel this downside, Chris Ré and others have introduced HipKittens, a library of programming primitive operations for AMD GPUs.
- OpenAI has launched GPT-5.1. The 2 new fashions are Instantaneous, which is tuned to be extra conversational and “human,” and Pondering, a reasoning mannequin that now adapts the time it takes to “suppose” to the issue of the questions.
- Giant language fashions, together with GPT-5 and the Chinese language fashions, present bias in opposition to customers who use a German dialect moderately than normal German. The bias seemed to be better because the mannequin dimension elevated. These outcomes additionally apply to languages like English.
- Ethan Mollick on evaluating (finally, interviewing) your AI fashions is a must-read.
- Yann LeCun is leaving Fb to launch a brand new startup that can develop his concepts about constructing AI.
- Harbor is a brand new device that simplifies benchmarking frameworks and fashions. It’s from the builders of the Terminal-Bench benchmark. And it brings us a step nearer to a world the place individuals construct their very own specialised AI moderately than depend on giant suppliers.
- Music rights holders are starting to make offers with Udio (and presumably different firms) that prepare their fashions on current music. Sadly, this doesn’t clear up the larger downside: Music is a “collectively produced shared cultural good, sustained by human labor. Copyright isn’t suited to defending this type of shared worth,” as professors Oliver Bown and Kathy Bowrey have argued.
- Moonshot AI has lastly launched Kimi K2 Pondering, the primary open weights mannequin to have benchmark outcomes aggressive with—or exceeding—the most effective closed weights fashions. It’s designed for use as an agent, calling exterior instruments as wanted to resolve issues.
- Tongyi DeepResearch is a brand new totally open supply agent for doing analysis. Its outcomes are corresponding to OpenAI deep analysis, Claude Sonnet 4, and comparable fashions. Tongyi is a part of Alibaba; it’s yet one more essential mannequin to return out of China.
- Knowledge facilities in house? It’s an attention-grabbing and difficult concept. Cooling is a a lot greater downside than you’d anticipate. They’d require huge arrays of photo voltaic cells for energy. However some individuals suppose it’d occur.
- MiniMax M2 is a brand new open weights mannequin that focuses on constructing brokers. It has efficiency much like Claude Sonnet however at a a lot lower cost level. It additionally embeds its thought processes between <suppose> and </suppose> tags, which is a vital step towards interpretability.
- DeepSeek has launched a new mannequin for OCR with some very attention-grabbing properties: It has a brand new course of for storing and retrieving recollections that additionally makes the mannequin considerably extra environment friendly.
- Agent Lightning gives a code-free approach to prepare brokers utilizing reinforcement studying.
Programming
- The Zig programming language has printed a ebook. On-line, after all.
- Google is weakening its controversial new guidelines about developer verification. The corporate plans to create a separate class for functions with restricted distribution, and develop a circulation that can enable the set up of unverified apps.
- Google’s LiteRT is a library for working AI fashions in browsers and small units. LiteRT helps Android, iOS, embedded Linux, and microcontrollers. Supported languages embrace Java, Kotlin, Swift, Embedded C, and C++.
- Does AI-assisted coding imply the top of latest languages? Simon Willison thinks that LLMs can encourage the event of latest programming languages. Design your language and ship it with a Claude Abilities-style doc; that needs to be sufficient for an LLM to learn to use it.
- Deepnote, a successor to the Jupyter Pocket book, is a next-generation pocket book for information analytics that’s constructed for groups. There’s now a shared workspace; totally different blocks can use totally different languages; and AI integration is on the highway map. It’s now open supply.
- The thought of assigning colours (crimson, blue) to instruments could also be useful in limiting the chance of immediate injection when constructing brokers. What instruments can return one thing damaging? This feels like a step in direction of the applying of the “least privilege” precept to AI design.
Safety
- We’re making the identical mistake with AI safety as we made with cloud safety (and safety normally): treating safety as an afterthought.
- Anthropic claims to have disrupted a Chinese language cyberespionage group that was utilizing Claude to generate assaults in opposition to different methods. Anthropic claims that the assault was 90% automated, although that declare is controversial.
- Don’t develop into a sufferer. Knowledge collected for on-line age verification makes your website a goal for attackers. That information is efficacious, they usually comprehend it.
- A analysis collaboration makes use of information poisoning and AI to disrupt deepfake pictures. Customers use Silverer to course of their pictures earlier than posting. The device makes invisible modifications to the unique picture that confuse AIs creating new pictures, resulting in unusable distortions.
- Is it a shock that AI is getting used to generate pretend receipts and expense experiences? In spite of everything, it’s used to pretend nearly all the things else. It was inevitable that enterprise functions of AI fakery would seem.
- HydraPWK2 is a Linux distribution designed for penetration testing. It’s primarily based on Debian and is supposedly simpler to make use of than Kali Linux.
- How safe is your trusted execution setting (TEE)? All the main {hardware} distributors are susceptible to numerous bodily assaults in opposition to “safe enclaves.” And their phrases of service usually exclude bodily assaults.
- Atroposia is a new malware-as-a-service bundle that features a native vulnerability scanner. As soon as an attacker has damaged right into a website, they will discover different methods to stay there.
- A brand new type of phishing assault (CoPhishing) makes use of Microsoft Copilot Studio brokers to steal credentials by abusing the Signal In subject. Microsoft has promised an replace that can defend in opposition to this assault.
Operations
- Right here’s the right way to set up Open Pocket book, an open supply equal to NotebookLM, to run by yourself {hardware}. It makes use of Docker and Ollama to run the pocket book and the mannequin regionally, so information by no means leaves your system.
- Open supply isn’t “free as in beer.” Neither is it “free as in freedom.” It’s “free as in puppies.” For higher or for worse, that almost says it.
- Want a framework for constructing proxies? Cloudflare’s subsequent technology Oxy framework could be what you want. (No matter you consider their latest misadventure.)
- MIT Media Labs’ Undertaking NANDA intends to construct infrastructure for a decentralized community of AI brokers. They describe it as a world decentralized registry (not in contrast to DNS) that can be utilized to find and authenticate brokers utilizing MCP and A2A. Isn’t this what we needed from the web within the first place?
Net
- The unfold of misinformation will be lowered, although not stopped, by including a small quantity of friction to the sharing course of.
- Netflix has launched a wonderful information to utilizing generative AI to create content material.
- Luke Wroblewski suggests a new mannequin for designing AI chat periods. A easy chat isn’t so simple as it appears; significantly with reasoning fashions, it may develop into cluttered to the purpose of uselessness. This new design addresses these issues.
Issues
- You want the Barely Annoying Rubik’s Dice Computerized Fixing Machine aka S.A.R.C.A.S.M. And you may make your individual.
- Monitoring butterfly migration: Are you able to make a sensor with related telemetry sufficiently small to suit on a butterfly tag? Sure, you’ll be able to!