Sabitlenmiş Tweet
Wes Roth
15.1K posts
Wes Roth
@WesRoth
FOLLOWS YOU. Artificial Intelligence, Automation & Optimism. Everything I say is 100% serious...
San Diego, CA Katılım Temmuz 2022
822 Takip Edilen35.5K Takipçiler
@deanwball anthropic explains what they mean by "functional feelings". It's something that mirrors human emotions to better predict the next token.
I don't think they mean it's "experiencing" anything
here's the relevant paper:
anthropic.com/research/emoti…
English
This, from Olah, plainly contradicts the encyclical, which confidently asserts that AI does not have, and never will have, “real” thoughts or feelings. It’s disappointing to see Anthropic align itself with a document that violates their own moral and intellectual principles.
English
xAI released Grok Build Beta for all SuperGrok and X Premium+ users, expanding access to its agentic building tool.
Grok Build now lets users work with Plan Mode, create images and videos through Imagine, and build automations or orchestrators using the CLI.
xAI@xai
Grok Build is now available in Beta for all SuperGrok and X Premium+ users. Use Plan Mode, create images and videos with Imagine, and build automations or orchestrators with the CLI. Visit x.ai/cli to get started.
English
Google Antigravity released the Antigravity CLI, a lightweight command-line tool for running the same Antigravity agents directly from the terminal.
The CLI gives developers access to the same harness and same models used in Antigravity, but with an experience built for terminal workflows.
Google Antigravity@antigravity
We’re thrilled to announce the Antigravity CLI, a lightweight way to spin up the same Antigravity agents right from the terminal. 💻 It gives you the exact same harness and same models, with a product experience tailored for the command line. It adapts entirely to you: your keybindings, your themes, your workflows. Full Antigravity CLI Walkthrough:
English
Google added a Gemini 3.5 Flash Low option to Antigravity, giving users a lighter model setting with lower token consumption.
Varun Mohan@_mohansolo
We heard concerns that Antigravity consumes many tokens for simple tasks now. So, we're adding Gemini 3.5 Flash (Low) as a way to optimize token usage for these tasks. In our internal testing, it generates around 45% fewer tokens than Gemini 3.5 Flash (Medium) and generally outperforms Gemini 3 Flash (High) on SWE tasks. We've also gone ahead and reset Gemini quota across all paid plans to make sure you have all the tokens needed to build for the next week 🙂
English
Wes Roth retweetledi
Grok memory is expected to become more customizable in the future, giving users more control over what the AI remembers.
Users will reportedly be able to view a summary of their existing memories and modify them directly.
️️️️ ️ᅠ️️️️ ️ᅠ️️️️ ️️️️️ ️ᅠ@blankspeaker
BREAKING: Here is a sneak peek at Grok's new Memory Settings. Grok remembers everything about you now and you can view and edit the memory file it keeps on you. This makes future interactions go much smoother as it has the added context.
English
Wes Roth retweetledi
Anthropic is reportedly preparing a new file-based memory upgrade for Claude, giving users the option to choose between Memory Files and Classic memory.
Memory Files are described as organized notes that Claude writes while users chat and reads back when they become relevant.
🚨 AI News | TestingCatalog@testingcatalog
ANTHROPIC 🔥: Claude will soon receive a new file-based memory upgrade, offering users the option to choose between Memory Files and Classic memory. > Organized notes Claude writes as you chat and reads when they're relevant. Browse and edit them anytime. This feature appears to be a new iteration of the previously discovered "Knowledge Bases" and more closely resembles what memory works in always-on agents like OpenClaw and Hermes. Considering a potential future debut of Claude Conway, Memory Files feature is likely an important preparation step.
English
Wes Roth retweetledi
Grok Build is reportedly being prepared for rollout to SuperGrok users, expanding access beyond the current SuperGrok Heavy tier.
Grok Build is currently available as a command-line tool, with a desktop app expected later.
Nima Owji@nima_owji
🚨 "SuperGrok" subscribers will get access to Grok Build (Beta) soon! (Currently, only the "SuperGrok Heavy" subscribers can access it!)
English
Wes Roth retweetledi
DeepSeek permanently reduced pricing for DeepSeek V4 Pro by 75%, making the model significantly cheaper to use.
The new pricing puts DeepSeek V4 Pro at $0.435 per million input tokens and $0.87 per million output tokens.
DeepSeek@deepseek_ai
We are making our discount permanent! 🎉 Enjoy building with DeepSeek-V4-Pro and bring your innovative ideas to life! 🚀
English
Claude is reportedly preparing a new file-based memory upgrade that would let users choose between Memory Files and Classic memory.
Memory Files are described as organized notes that Claude writes during conversations and reads later when relevant.
🚨 AI News | TestingCatalog@testingcatalog
ANTHROPIC 🔥: Claude will soon receive a new file-based memory upgrade, offering users the option to choose between Memory Files and Classic memory. > Organized notes Claude writes as you chat and reads when they're relevant. Browse and edit them anytime. This feature appears to be a new iteration of the previously discovered "Knowledge Bases" and more closely resembles what memory works in always-on agents like OpenClaw and Hermes. Considering a potential future debut of Claude Conway, Memory Files feature is likely an important preparation step.
English
If you can’t keep shrinking transistors forever, you need a new way to keep improving chip performance.
Huawei@Huawei
HUAWEI's Tau (τ) Scaling Law is a new principle for guiding the future development of semiconductors. By 2031, HUAWEI's high-end chips are expected to feature a transistor density equivalent to 14 Å (1.4 nm) processes. Watch the livestream to learn more! twitter.com/i/broadcasts/1…
English
Grok foundation model V9-Medium, a 1.5T-parameter model, has reportedly finished training and is now moving toward public release.
The model is expected to be a major upgrade over the current 0.5T V8-Small model that serves Grok production traffic.
Elon Musk@elonmusk
Grok foundation model V9-Medium (1.5T) has finished training. Evals look good. A lot of Cursor data was added in supplementary training and there is more to come. Fine-tuning is underway and reinforcement learning begins in a few days. 2 to 3 weeks to public release. This will be a major improvement over the 0.5T v8-small that currently serves all Grok production traffic, especially for difficult coding tasks.
English
Wes Roth retweetledi
OpenAI's Daybreak is now being used by major Australian banks for cyber defence.
Commonwealth Bank and Westpac have turned to OpenAI’s model to test their cybersecurity systems for vulnerabilities.
The move comes after Anthropic restricted access to its rival cyber model, Mythos, to a limited group of US technology companies.
Britain’s AI Security Institute rated GPT-5.5-Cyber as almost as reliable as Anthropic’s Mythos at conducting complex cyber tasks.
English
Wes Roth retweetledi
Anthropic appears to be preparing Mythos 1, listed as “claude-mythos-1-preview,” for Claude Code and Claude Security.
The model was briefly visible on Claude, and new strings mentioning Mythos were reportedly added.
This is worth watching, but I’d be careful not to treat it like a full public launch yet.
🚨 AI News | TestingCatalog@testingcatalog
ANTHROPIC 🔥: Mythos 1, "claude-mythos-1-preview", is being prepared for a release on Claude Code and Claude Security. The model became visible for a short amount of time on Claude; besides that, new strings mentioning Mythos have been added. > Access to the Claude Mythos model in Claude Code and Claude Security. It still doesn't mean the general public will have access to this exact model, according to Anthropic's earlier communication. More below 👇
English
Wes Roth retweetledi
This is one of the more interesting AI math results I’ve seen recently.
DeepMind researchers published a new paper showing AI-driven formal proof search can solve real open mathematics problems.
The system autonomously resolved 9 out of 353 open Erdős problems using LLM agents and Lean-based formal verification.
Key details:
🔹The approach uses LLMs to generate formal proofs in Lean, reducing the reliability problem that comes with informal AI-generated math reasoning.
🔹The strongest agent solved 9 open Erdős problems at a cost of a few hundred dollars per problem.
🔹The system also proved 44 out of 492 OEIS conjectures.
🔹A simpler agent that alternated between LLM generation and Lean verification replicated the Erdős results, but was more expensive on the hardest problems.
🔹The workflow points to a powerful loop: AI agents search for proofs, Lean formally verifies them, and human researchers review the final results.
🔹The system is already being deployed in areas like combinatorics, optimization, graph theory, algebraic geometry, and quantum optics.
English
Wes Roth retweetledi
Anthropic and its partners have found more than 10,000 high- or critical-severity vulnerabilities in essential software since launching last month.
The initiative uses advanced AI cybersecurity capabilities, including models like Claude Mythos Preview, to identify serious vulnerabilities.
Anthropic@AnthropicAI
Last month we launched Project Glasswing, our collaborative AI cybersecurity initiative. Since then, we and our partners have found more than ten thousand high- or critical-severity vulnerabilities in essential software.
English
Wes Roth retweetledi
Anthropic appears to be signaling that Mythos-class cybersecurity models may eventually get a broader general release, but only after stronger safeguards are developed.
Andrew Curran@AndrewCurran_
Looks like they meant it.
English
Google launched a free Google AI Studio feature that lets users build native Android apps directly inside AI Studio without coding.
The feature has already been used to create more than 250,000 Android apps since launching last week.
Logan Kilpatrick@OfficialLoganK
We just launched the ability to build native Android apps directly in Google AI Studio for free! Since launch last week, people have created more than 250,000 Android apps. Likely >99% of these folks never built an Android app before, everyone can now build, no coding required!
English
Pope Leo XIV announced a collaboration with Anthropic focused on AI ethics and humanity’s shared future.
He released his first encyclical, Magnifica Humanitas, focused on the risks artificial intelligence poses to human dignity, inequality, warfare, and social stability.
The document frames unchecked AI as a modern “Tower of Babel,” warning that the technology can reflect human greed and the pursuit of absolute power instead of shared benefit.
English
Moore’s Law is getting harder, slower, and more expensive.
Huawei presented the Tau Scaling Law, a new semiconductor principle designed to guide chip progress as Moore’s Law faces physical and economic limits.
The Tau Scaling Law shifts the focus from geometric transistor shrinking to reducing signal propagation delay, or time constant τ, across devices, circuits, chips, and systems.
The company says traditional transistor scaling is slowing down, making it harder to keep improving performance and cost efficiency through smaller process nodes alone.
Tau Scaling focuses on shortening signal delay to improve performance, energy efficiency, and transistor density.
Huawei’s LogicFolding architecture is a core technology under this approach, designed to shorten critical-path wiring and improve circuit performance.
At the chip level, Huawei is using full-stack software, architecture, and silicon co-design to improve instruction and data flow.
At the system level, Huawei is using UnifiedBus to reduce communication latency across large computing systems.
Huawei says it has designed and mass-produced 381 chips based on the Tau Scaling Law over the past six years.
Kirin chips launching in Fall 2026 are expected to be the first to use the LogicFolding architecture.
Huawei projects that by 2031, its high-end chips based on Tau Scaling could reach transistor density equivalent to 14 Å, or 1.4 nm, processes.
Huawei@Huawei
HUAWEI has presented the Tau (τ) Scaling Law, a new principle for guiding the future development of the semiconductor industry. By 2031, HUAWEI's high-end chips based on this law are expected to feature a transistor density that is equivalent to 14 Å (1.4 nm) processes.
English
Perplexity open-sourced Bumblebee, a read-only security scanner for macOS and Linux developer machines.
Bumblebee checks for risky packages, browser or editor extensions, and AI tool configuration files that could expose developers to supply-chain attacks.
Perplexity@perplexity_ai
Today we're open-sourcing Bumblebee, a read-only scanner for macOS and Linux. It checks developer machines for risky packages, extensions, and AI tool configs. Connected to Computer, it can trigger deeper scans whenever a new supply-chain risk emerges. github.com/perplexityai/b…
English