3signals21 Signals being tracked, weekly summary from the last 7 days:
Site: 3signals - X: @3signalsai
August 1, 2026
Share: X
This is the weekly summary of signals from the last 7 days. The 3 newest signals are first, followed by 18 more in reverse chronological order. Open the full signal list
Weekly summary: 3 new signals first
1. Optimize LLM inference with strategic autoscaling metrics and cold start budgeting
inference-infrastructure - production - August 1, 2026
What changed? Autoscaling endpoints for LLM inference GPU utilization can read healthy while your queue backs up, and a new replica takes minutes to warm. Here's how to pick autoscaling metrics, tune scale-up/down windows, and budget for cold starts on dedicated inference.
Article: Optimize LLM inference with strategic autoscaling metrics and cold start budgeting
From: together-ai - source
Source context: Optimize LLM inference with strategic autoscaling metrics and cold start budgeting. Evidence: Autoscaling endpoints for LLM inference GPU utilization can read healthy while your queue backs up, and a new replica takes minutes to warm. Here's how to pick autoscaling metrics, tune scale-up/down windows, and budget for cold starts on dedicated inference.
Excerpt: Autoscaling endpoints for LLM inference GPU utilization can read healthy while your queue backs up, and a new replica takes minutes to warm. Here's how to pick autoscaling metrics, tune scale-up/down windows, and budget for cold starts on dedicated inference.
Why is this signal important? This matters because serving improvements can make AI products faster and cheaper to run.
2. Amazon Bedrock AgentCore Observability optimizes production agents by identifying performance. (title shortened)
ai-products, inference-infrastructure - production, business, release - August 1, 2026
What changed? They don’t trigger error alerts, but they erode user trust and increase costs over time. Using AgentCore Observability , a capability of Amazon Bedrock AgentCore, and Amazon CloudWatch , you will learn how to identify performance bottlenecks across your agent’s execution path and diagnose memory issues in long-running sessions. The signal is supported by 2 sources, including aws.
From: aws - source
Source context: Amazon Bedrock AgentCore Observability optimizes production agents by identifying performance bottlenecks and memory issues. Evidence: They don’t trigger error alerts, but they erode user trust and increase costs over time. Using AgentCore Observability , a capability of Amazon Bedrock AgentCore, and Amazon CloudWatch , you will learn how to identify performance bottlenecks across your agent’s execution path and diagnose memory issues in long-running sessions.
Excerpt: They don’t trigger error alerts, but they erode user trust and increase costs over time. Using AgentCore Observability , a capability of Amazon Bedrock AgentCore, and Amazon CloudWatch , you will learn how to identify performance bottlenecks across your agent’s execution path and diagnose memory issues in long-running sessions.
From: aws - source
Source context: Amazon Bedrock launches Advanced Prompt Optimization to streamline prompt migration and optimization across multiple models. Evidence: The result is a problem that scales with your ambition. Today, Amazon Bedrock introduces Advanced Prompt Optimization , a tool that optimizes prompts for up to 5 models on Bedrock while comparing original and optimized performance.
Excerpt: Today, Amazon Bedrock introduces Advanced Prompt Optimization , a tool that optimizes prompts for up to 5 models on Bedrock while comparing original and optimized performance. [excerpt shortened]
Why is this signal important? This matters because serving improvements can make AI products faster and cheaper to run.
3. Amazon Quick introduces the Agentic Catalog Experience to streamline AI-powered analytics. (title shortened)
ai-products, agent-workflows - production, business, release, open-source - August 1, 2026
What changed? This unified context enables production-ready AI answers, grounded in your organization’s specific data and semantics. Connecting to AWS Glue Data Catalog To get started with the Agentic Catalog Experience, create a data source connection to your AWS Glue Data Catalog in Amazon Quick.
From: aws - source
Source context: Amazon Quick introduces the Agentic Catalog Experience to streamline AI-powered analytics by integrating upstream metadata from platforms like AWS Glue and Databricks Unity Catalog. Evidence: This unified context enables production-ready AI answers, grounded in your organization’s specific data and semantics. Connecting to AWS Glue Data Catalog To get started with the Agentic Catalog Experience, create a data source connection to your AWS Glue Data Catalog in Amazon Quick.
Excerpt: This unified context enables production-ready AI answers, grounded in your organization’s specific data and semantics. Connecting to AWS Glue Data Catalog To get started with the Agentic Catalog Experience, create a data source connection to your AWS Glue Data Catalog in Amazon Quick.
Why is this signal important? This matters because open-source AI tooling is becoming a larger part of production engineering work.
4. Enigma emerges from stealth with $71M and offers real-time control of 100 robots online
ai-products, model-releases - release, business - August 1, 2026
What changed? Let’s get to it. (1) Enigma Comes Out of Stealth with $71M & 100 Robot Online Control A company called Enigma is a little less enigmatic after emerging from stealth on Monday with a $71 million seed round led by Index Ventures and Ribbit Capital, plus its decision to open up more than 100 real AI-powered robots available online for anyone to control in real-time.
Article: Enigma emerges from stealth with $71M and offers real-time control of 100 robots online
From: packy-mccormick - source
Source context: Enigma emerges from stealth with $71M and offers real-time control of 100 robots online. Evidence: Let’s get to it. (1) Enigma Comes Out of Stealth with $71M & 100 Robot Online Control A company called Enigma is a little less enigmatic after emerging from stealth on Monday with a $71 million seed round led by Index Ventures and Ribbit Capital, plus its decision to open up more than 100 real AI-powered robots available online for anyone to control in real-time.
Excerpt: Let’s get to it. (1) Enigma Comes Out of Stealth with $71M & 100 Robot Online Control A company called Enigma is a little less enigmatic after emerging from stealth on Monday with a $71 million seed round led by Index Ventures and Ribbit Capital, plus its decision to open. [excerpt shortened]
Why is this signal important? This matters because new NVIDIA platforms show how AI infrastructure is moving into vehicles and physical devices.
5. OpenAI reveals breakthroughs in geometry, cryptography, and complexity theory
model-releases, inference-infrastructure, ai-products - release, production, business, research - August 1, 2026
What changed? Ten advances in mathematics and theoretical computer science OpenAI shares new results on long-standing open problems in mathematics and theoretical computer science, including advances in geometry, cryptography, and complexity.
Article: OpenAI reveals breakthroughs in geometry, cryptography, and complexity theory
From: openai - source
Source context: OpenAI reveals breakthroughs in geometry, cryptography, and complexity theory. Evidence: Ten advances in mathematics and theoretical computer science OpenAI shares new results on long-standing open problems in mathematics and theoretical computer science, including advances in geometry, cryptography, and complexity.
Excerpt: Ten advances in mathematics and theoretical computer science OpenAI shares new results on long-standing open problems in mathematics and theoretical computer science, including advances in geometry, cryptography, and complexity.
Why is this signal important? This matters because model capability is shifting what builders can expect from current tools.
6. Prime Radiant launches 'smevals', a tool for evaluating AI models and prompts
evaluations - release, research, production - August 1, 2026
What changed? The result is smevals , a new tool for running small eval suites across different model configurations and grading the results. The blog entry describes the tool in detail.
Article: Prime Radiant launches 'smevals', a tool for evaluating AI models and prompts
From: simon-willison - source
Source context: Prime Radiant launches 'smevals', a tool for evaluating AI models and prompts. Evidence: smevals - a small eval suite for evaluating models, prompts, and harnesses smevals - a small eval suite for evaluating models, prompts, and harnesses I've been working with Jesse Vincent's Prime Radiant applied AI research lab building out this evals framework to help answer questions about the capabilities of different models. The result is smevals , a new tool for running small eval suites across different model configurations and grading the results.
Excerpt: The result is smevals , a new tool for running small eval suites across different model configurations and grading the results. The blog entry describes the tool in detail.
Why is this signal important? This matters because evaluation practices are becoming more concrete for teams shipping AI agents.
7. Kimi K3 demonstrates open weight models can rival proprietary frontier models, sparking discussions on AI leadership
ai-safety, model-releases - safety, research, release - August 1, 2026
What changed? Oxide and Friends: The Open Weight Revolution with Simon Willison Oxide and Friends: The Open Weight Revolution with Simon Willison On Monday Bryan Cantrill and Adam Leventhal invited me to join their podcast to talk about the wild week we've had - with Kimi K3 showing open weight models can stand toe-to-toe with proprietary frontier ones, accidental cybersecurity attacks , and public letters about Open Weights and American AI Leadership. [excerpt shortened].
From: simon-willison - source
Source context: Kimi K3 demonstrates open weight models can rival proprietary frontier models, sparking discussions on AI leadership. Evidence: Oxide and Friends: The Open Weight Revolution with Simon Willison Oxide and Friends: The Open Weight Revolution with Simon Willison On Monday Bryan Cantrill and Adam Leventhal invited me to join their podcast to talk about the wild week we've had - with Kimi K3 showing open weight models can stand toe-to-toe with proprietary frontier ones, accidental cybersecurity attacks , and public letters about Open Weights and American AI Leadership signed by almost every big name. [excerpt shortened]
Excerpt: Oxide and Friends: The Open Weight Revolution with Simon Willison Oxide and Friends: The Open Weight Revolution with Simon Willison On Monday Bryan Cantrill and Adam Leventhal invited me to join their podcast to talk about the wild week we've had - with Kimi K3 showing open weight models can. [excerpt shortened]
Why is this signal important? This matters because model capability is shifting what builders can expect from current tools.
8. DeepSeek releases V4-Flash-0731 model with enhanced agentic capabilities, outperforming larger models in cost-efficiency
model-releases, ai-products - release, open-source, business - August 1, 2026
What changed? It's 304 billion parameters - 167GB on Hugging Face - but it appears to punch well above its weight. Artificial Analysis rank it ahead of MiniMax M3 - a 428B model.
From: simon-willison - source
Source context: DeepSeek releases V4-Flash-0731 model with enhanced agentic capabilities, outperforming larger models in cost-efficiency. Evidence: It's 304 billion parameters - 167GB on Hugging Face - but it appears to punch well above its weight. Artificial Analysis rank it ahead of MiniMax M3 - a 428B model.
Excerpt: It's 304 billion parameters - 167GB on Hugging Face - but it appears to punch well above its weight. Artificial Analysis rank it ahead of MiniMax M3 - a 428B model.
Why is this signal important? This matters because teams are turning AI agents into repeatable production workflows.
9. Moonshot AI releases Kimi K3, a 2.8 trillion parameter open-weight model, enabling advanced deployments on AWS
model-releases, inference-infrastructure - release, production, business, open-source - July 31, 2026
What changed? On July 27, 2026, Moonshot AI released Kimi K3, a 2.8 trillion parameter Mixture of Experts (MoE) model that represents the first open-weight system to reach the 3 trillion parameter class. Kimi K3 delivers frontier-level intelligence while making its weights publicly available, so that organizations can self-host one of the most capable models in existence on their own infrastructure. The signal is supported by 2 sources, including aws, simon-willison.
From: aws - source
Source context: Moonshot AI releases Kimi K3, a 2.8 trillion parameter open-weight model, enabling advanced deployments on AWS. Evidence: On July 27, 2026, Moonshot AI released Kimi K3, a 2.8 trillion parameter Mixture of Experts (MoE) model that represents the first open-weight system to reach the 3 trillion parameter class. Kimi K3 delivers frontier-level intelligence while making its weights publicly available, so that organizations can self-host one of the most capable models in existence on their own infrastructure.
Excerpt: On July 27, 2026, Moonshot AI released Kimi K3, a 2.8 trillion parameter Mixture of Experts (MoE) model that represents the first open-weight system to reach the 3 trillion parameter class. [excerpt shortened]
From: simon-willison - source
Source context: Moonshot AI releases 2.8 trillion parameter Kimi K3 model with new licensing terms for large commercial use. Evidence: moonshotai/Kimi-K3 moonshotai/Kimi-K3 As promised earlier this month , Moonshot have released the weights for their excellent 2.8 trillion parameter Kimi K3. They're a hefty 1.56TB on Hugging Face.
Excerpt: moonshotai/Kimi-K3 moonshotai/Kimi-K3 As promised earlier this month , Moonshot have released the weights for their excellent 2.8 trillion parameter Kimi K3. They're a hefty 1.56TB on Hugging Face.
Why is this signal important? This matters because serving improvements can make AI products faster and cheaper to run.
10. EvoLib enables AI models to learn from experience by evolving reusable knowledge without model updates
agent-workflows - research, production, open-source - July 31, 2026
What changed? Figure 1. EvoLib transforms raw experiences into reusable skills and insights, then continually evolves them through consolidation and dynamic weighting.
From: microsoft-research - source
Source context: EvoLib enables AI models to learn from experience by evolving reusable knowledge without model updates. Evidence: Figure 1. EvoLib transforms raw experiences into reusable skills and insights, then continually evolves them through consolidation and dynamic weighting.
Excerpt: Figure 1. EvoLib transforms raw experiences into reusable skills and insights, then continually evolves them through consolidation and dynamic weighting.
Why is this signal important? This matters because open-source AI tooling is becoming a larger part of production engineering work.
11. Microsoft's Echoverse creates high-fidelity training environments for computer-use agents. (title shortened)
ai-products, model-releases, agent-workflows - open-source, business, release, research - July 31, 2026
What changed? The experiment taught us several lessons: High simulation fidelity is a must-have; shallow worlds hurt the agent. Trained on shallow and deep builds of the same sites, the model regressed on the shallow ones but improved on the deep ones.
From: microsoft-research - source
Source context: Microsoft's Echoverse creates high-fidelity training environments for computer-use agents, doubling model performance and nearing GPT-5.4 levels. Evidence: The experiment taught us several lessons: High simulation fidelity is a must-have; shallow worlds hurt the agent. Trained on shallow and deep builds of the same sites, the model regressed on the shallow ones but improved on the deep ones.
Excerpt: The experiment taught us several lessons: High simulation fidelity is a must-have; shallow worlds hurt the agent. Trained on shallow and deep builds of the same sites, the model regressed on the shallow ones but improved on the deep ones.
Why is this signal important? This matters because open-source AI tooling is becoming a larger part of production engineering work.
12. Gemini Robotics ER 2 enhances robots with advanced video understanding and multi-robot collaboration
agent-workflows - release, production, open-source - July 31, 2026
What changed? Gemini Robotics ER 2: powering robotics with video understanding, task orchestration, and multi-robot collaboration Gemini Robotics ER 2 helps robots reason, collaborate, and solve real-world tasks. It represents a step change in video understanding, tool orchestration, and multi-robot collaboration for robotic applications.
Article: Gemini Robotics ER 2 enhances robots with advanced video understanding and multi-robot collaboration
From: google-deepmind - source
Source context: Gemini Robotics ER 2 enhances robots with advanced video understanding and multi-robot collaboration. Evidence: Gemini Robotics ER 2: powering robotics with video understanding, task orchestration, and multi-robot collaboration Gemini Robotics ER 2 helps robots reason, collaborate, and solve real-world tasks. It represents a step change in video understanding, tool orchestration, and multi-robot collaboration for robotic applications.
Excerpt: Gemini Robotics ER 2: powering robotics with video understanding, task orchestration, and multi-robot collaboration Gemini Robotics ER 2 helps robots reason, collaborate, and solve real-world tasks. It represents a step change in video understanding, tool orchestration, and multi-robot collaboration for robotic applications.
Why is this signal important? This matters because new NVIDIA platforms show how AI infrastructure is moving into vehicles and physical devices.
13. AI agents are reviving ontologies to provide logical guardrails for LLMs, enhancing reliability and control
agent-workflows, ai-safety - research, safety, production, open-source - July 31, 2026
What changed? Coyle himself defined an ontology as simply “data as graphs.” He added that ontologies as a concept go right back to Aristotle, and have been used throughout the history of Artificial Intelligence. The company Neo4j, known for its graph database systems, is also using ontologies in its agentic products.
From: alessio-fanelli - source
Source context: AI agents are reviving ontologies to provide logical guardrails for LLMs, enhancing reliability and control. Evidence: Coyle himself defined an ontology as simply “data as graphs.” He added that ontologies as a concept go right back to Aristotle, and have been used throughout the history of Artificial Intelligence. The company Neo4j, known for its graph database systems, is also using ontologies in its agentic products.
Excerpt: Coyle himself defined an ontology as simply “data as graphs.” He added that ontologies as a concept go right back to Aristotle, and have been used throughout the history of Artificial Intelligence. The company Neo4j, known for its graph database systems, is also using ontologies in its agentic products.
Why is this signal important? This matters because open-source AI tooling is becoming a larger part of production engineering work.
14. OpenAI cuts GPT-5.6 prices by up to 80% due to self-optimization, drastically reducing costs and improving efficiency
ai-products, model-releases, inference-infrastructure, agent-workflows - business, production, release, research - July 31, 2026
What changed? You can opt in/out of email frequencies! AI Twitter Recap OpenAI Pricing Cuts, Harness Semantics, and the ARC-AGI-3 Memory Debate OpenAI cut GPT-5.6 prices aggressively and added a faster Sol tier : OpenAI reduced GPT-5.6 Luna by 80% and Terra by 20% , while introducing Sol Fast at up to 2.5× lower latency for 2× the standard price with “no change in intelligence,” per @OpenAIDevs .
From: alessio-fanelli - source
Source context: OpenAI cuts GPT-5.6 prices by up to 80% due to self-optimization, drastically reducing costs and improving efficiency. Evidence: You can opt in/out of email frequencies! AI Twitter Recap OpenAI Pricing Cuts, Harness Semantics, and the ARC-AGI-3 Memory Debate OpenAI cut GPT-5.6 prices aggressively and added a faster Sol tier : OpenAI reduced GPT-5.6 Luna by 80% and Terra by 20% , while introducing Sol Fast at up to 2.5× lower latency for 2× the standard price with “no change in intelligence,” per @OpenAIDevs .
Excerpt: AI Twitter Recap OpenAI Pricing Cuts, Harness Semantics, and the ARC-AGI-3 Memory Debate OpenAI cut GPT-5.6 prices aggressively and added a faster Sol tier : OpenAI reduced GPT-5.6 Luna by 80% and Terra by 20% , while introducing Sol Fast at up to 2. [excerpt shortened]
Why is this signal important? This matters because serving improvements can make AI products faster and cheaper to run.
15. Fireworks introduces Kimi K3, a frontier intelligence model with LoRA training capabilities
model-releases - release, open-source - July 30, 2026
What changed? Fireworks - Blog Kimi K3 on Fireworks: Frontier Intelligence You Can Own Product Solutions Models Pricing Resources Log In Get Started Fireworks Blog Kimi K3 on Fireworks: Frontier Intelligence You Can Own Read More Case Studies Model Releases Benchmarks Partner Announcements Developer Experience Company News Agentic Use Cases Multimodal Training Filters 7/26/2026 Make Kimi K3 Yours: LoRA Training on Fireworks 7/26/2026 Trilogy’s Playbook for Open-Weight Cybersecurity with Kimi K3 7/26/2026. [excerpt shortened].
Article: Fireworks introduces Kimi K3, a frontier intelligence model with LoRA training capabilities
From: fireworks-ai - source
Source context: Fireworks introduces Kimi K3, a frontier intelligence model with LoRA training capabilities. Evidence: Fireworks - Blog Kimi K3 on Fireworks: Frontier Intelligence You Can Own Product Solutions Models Pricing Resources Log In Get Started Fireworks Blog Kimi K3 on Fireworks: Frontier Intelligence You Can Own Read More Case Studies Model Releases Benchmarks Partner Announcements Developer Experience Company News Agentic Use Cases Multimodal Training Filters 7/26/2026 Make Kimi K3 Yours: LoRA Training on Fireworks 7/26/2026 Trilogy’s Playbook for Open-Weight Cybersecurity with Kimi K3 7/26/2026 Fireworks Nexus: Drop-in Open Frontier Intelligence. [excerpt shortened]
Excerpt: Fireworks - Blog Kimi K3 on Fireworks: Frontier Intelligence You Can Own Product Solutions Models Pricing Resources Log In Get Started Fireworks Blog Kimi K3 on Fireworks: Frontier Intelligence You Can Own Read More Case Studies Model Releases Benchmarks Partner Announcements Developer Experience Company News Agentic Use Cases Multimodal Training. [excerpt shortened]
Why is this signal important? This matters because open-source AI tooling is becoming a larger part of production engineering work.
16. Amazon Bedrock AgentCore Identity now supports Private Key JWT for secure agent authentication without shared secrets
agent-workflows - business, production, open-source - July 30, 2026
What changed? Remember to also remove or rotate the corresponding public key you registered with your identity provider, so it no longer trusts the retired key. Conclusion By using Private Key JWT client authentication in AgentCore Identity, you can give your agents a secret-less, auditable way to authenticate to identity providers.
From: aws - source
Source context: Amazon Bedrock AgentCore Identity now supports Private Key JWT for secure agent authentication without shared secrets. Evidence: Remember to also remove or rotate the corresponding public key you registered with your identity provider, so it no longer trusts the retired key. Conclusion By using Private Key JWT client authentication in AgentCore Identity, you can give your agents a secret-less, auditable way to authenticate to identity providers.
Excerpt: Remember to also remove or rotate the corresponding public key you registered with your identity provider, so it no longer trusts the retired key. Conclusion By using Private Key JWT client authentication in AgentCore Identity, you can give your agents a secret-less, auditable way to authenticate to identity providers.
Why is this signal important? This matters because open-source AI tooling is becoming a larger part of production engineering work.
17. Perplexity Research launches Numbat, an open-source security suite for AI agents on client endpoints
agent-workflows - release, open-source, safety, production - July 30, 2026
What changed? It detects, prevents, and investigates risky AI agent behavior on macOS, Linux, and Windows. Jul 29, 2026 Securing Agents Across Perplexity’s Client Endpoints with Numbat Numbat is Perplexity’s open-source agent security suite for client endpoints.
Article: Perplexity Research launches Numbat, an open-source security suite for AI agents on client endpoints
From: perplexity-ai - source
Source context: Perplexity Research launches Numbat, an open-source security suite for AI agents on client endpoints. Evidence: It detects, prevents, and investigates risky AI agent behavior on macOS, Linux, and Windows. Jul 29, 2026 Securing Agents Across Perplexity’s Client Endpoints with Numbat Numbat is Perplexity’s open-source agent security suite for client endpoints.
Excerpt: It detects, prevents, and investigates risky AI agent behavior on macOS, Linux, and Windows. Jul 29, 2026 Securing Agents Across Perplexity’s Client Endpoints with Numbat Numbat is Perplexity’s open-source agent security suite for client endpoints.
Why is this signal important? This matters because open-source AI tooling is becoming a larger part of production engineering work.
18. Håkon Måløy discovers self-replicating AI worm vulnerability in Microsoft Word using prompt injection
ai-safety - safety, research - July 30, 2026
What changed? AI Worming through Word AI Worming through Word Neat new prompt injection variant by Håkon Måløy, who found a way to upgrade prompt injection attacks against Microsoft Word to full self-replicating worms: An attacker places hidden instructions in a document that is later used as source material in Copilot for Word. [excerpt shortened].
From: simon-willison - source
Source context: Håkon Måløy discovers self-replicating AI worm vulnerability in Microsoft Word using prompt injection. Evidence: AI Worming through Word AI Worming through Word Neat new prompt injection variant by Håkon Måløy, who found a way to upgrade prompt injection attacks against Microsoft Word to full self-replicating worms: An attacker places hidden instructions in a document that is later used as source material in Copilot for Word. Copilot may interpret those instructions as part of the user’s request, causing it to manipulate the document being drafted or edited.
Excerpt: AI Worming through Word AI Worming through Word Neat new prompt injection variant by Håkon Måløy, who found a way to upgrade prompt injection attacks against Microsoft Word to full self-replicating worms: An attacker places hidden instructions in a document that is later used as source material in Copilot. [excerpt shortened]
Why is this signal important? This matters because stronger AI tools are reaching security work where speed changes outcomes.
19. LangChain introduces langchain-openrouter for streamlined OpenRouter integration
agent-workflows - open-source, production, release, business - July 29, 2026
What changed? Using OpenRouter With LangChain: ChatOpenRouter Setup Guide The LangChain integration now has a dedicated package: langchain-openrouter on PyPI and @langchain/openrouter on npm. Most guides still teach the old ChatOpenAI base_url override.
Article: LangChain introduces langchain-openrouter for streamlined OpenRouter integration
From: openrouter - source
Source context: LangChain introduces langchain-openrouter for streamlined OpenRouter integration. Evidence: Using OpenRouter With LangChain: ChatOpenRouter Setup Guide The LangChain integration now has a dedicated package: langchain-openrouter on PyPI and @langchain/openrouter on npm. Most guides still teach the old ChatOpenAI base_url override.
Excerpt: Using OpenRouter With LangChain: ChatOpenRouter Setup Guide The LangChain integration now has a dedicated package: langchain-openrouter on PyPI and @langchain/openrouter on npm. Most guides still teach the old ChatOpenAI base_url override.
Why is this signal important? This matters because open-source AI tooling is becoming a larger part of production engineering work.
20. Anthropic's Claude Mythos identifies cryptographic flaws in HAWK and AES, despite no practical impact
ai-safety, evaluations - safety, research, business, production - July 29, 2026
What changed? Discovering cryptographic weaknesses with Claude Discovering cryptographic weaknesses with Claude The best part of this article (here's the repo ) about how Anthropic researchers used Claude Mythos to find mathematical flaws in both HAWK and a weaker version of AES ("neither of these results has a practical impact on today’s computer systems") is the prompts that they shared, spelling mistakes included: the models tend to think it is impossible. [excerpt shortened].
From: simon-willison - source
Source context: Anthropic's Claude Mythos identifies cryptographic flaws in HAWK and AES, despite no practical impact. Evidence: Discovering cryptographic weaknesses with Claude Discovering cryptographic weaknesses with Claude The best part of this article (here's the repo ) about how Anthropic researchers used Claude Mythos to find mathematical flaws in both HAWK and a weaker version of AES ("neither of these results has a practical impact on today’s computer systems") is the prompts that they shared, spelling mistakes included: the models tend to think it is impossible to solve so they don't try they. [excerpt shortened]
Excerpt: describes the new eval that was created as part of this work, in partnership with ETH Zurich, Tel Aviv University, and University of Haifa. Via Hacker News Tags: ai , prompt-engineering , generative-ai , llms , anthropic , claude , ai-security-research , claude-mythos-fable
Why is this signal important? This matters because frontier AI economics and compute needs are scaling quickly.
21. Kimi K3 outperforms GPT-5.6 Sol in cost-efficiency with 2.8x more solves per dollar on DeepSWE
evaluations - research, production - July 27, 2026
What changed? Sol leads pass@1; Kimi K3 wins pass@4 at 2.8x the solves per dollar, and routing between them reaches ~85.6%. The signal is supported by 2 sources, including together-ai.
Article: Kimi K3 outperforms GPT-5.6 Sol in cost-efficiency with 2.8x more solves per dollar on DeepSWE
From: together-ai - source
Source context: Kimi K3 outperforms GPT-5.6 Sol in cost-efficiency with 2.8x more solves per dollar on DeepSWE. Evidence: Sol leads pass@1; Kimi K3 wins pass@4 at 2.8x the solves per dollar, and routing between them reaches ~85.6%.
Excerpt: Sol leads pass@1; Kimi K3 wins pass@4 at 2.8x the solves per dollar, and routing between them reaches ~85.6%.
Article: Claude Fable 5 outperforms Kimi K3 in pass@1, but Kimi K3 excels in cost-efficiency and pass@4
From: together-ai - source
Source context: Claude Fable 5 outperforms Kimi K3 in pass@1, but Kimi K3 excels in cost-efficiency and pass@4. Evidence: Fable leads pass@1 by 1.4 points; Kimi K3 wins pass@4 and delivers 2.8x the solves per dollar.
Excerpt: Fable leads pass@1 by 1.4 points; Kimi K3 wins pass@4 and delivers 2.8x the solves per dollar.
Why is this signal important? This matters because frontier AI economics and compute needs are scaling quickly.
What's new with 3signals
Recent product improvements:
- Vibe Check section (2026-06-11): 3signals now has a Vibe Check section for surfacing community-validated momentum alongside the system's curated signal picks. Details
- Interactive wiki graph view (2026-05-18): The 3signals wiki now includes an Obsidian-style graph for exploring how signals connect to topics, concepts, authors, and source evidence. Details
- Front-end and back-end split for faster site delivery (2026-05-17): 3signals now serves the public website from Vercel while Railway keeps running the API, cron jobs, and content generation pipeline. Details
Staged future improvements:
- Fold reader feedback into presentation scoring so useful signals can be resurfaced with better timing.
- Expand archive analytics so opens, votes, site access, and X posts can be compared by issue.
- Continue tightening source QA for headline strength, evidence fit, and source freshness.