absorb.md

swyx

Chronological feed of everything captured from swyx.

AI Engineer Europe 2025: Harness Engineering, Open Claude Security, and the Claude Mythos Preview

Recorded live from AI Engineer Europe in London, this episode covers the week's major AI developments: Anthropic's preview release of Claude Mythos (restricted to ~40 enterprise partners under Project Glass Wing due to unprecedented cybersecurity capabilities), the rapid growth of agentic coding tools like Open Claude and Codex, and the emerging consensus around "harness engineering" and context engineering as durable pillars of agent-based AI. Swyx and other practitioners argue that extreme parallelism in multi-agent systems is placing unprecedented load on infrastructure like GitHub, while security concerns around supply-chain attacks (LiteLLM, Axios, Open Claude) are pushing developers toward zero-dependency, self-owned codebases. The episode also highlights that Anthropic's decision not to release Mythos publicly may be as much about compute constraints ($30B ARR but limited GPU capacity) as safety posture.

Anthropic’s Mythos Model: A Deep Dive into its Capabilities, Security Concerns, and Market Impact

Anthropic's recent Mythos model, despite not being publicly released due to perceived security risks, showcases significant advancements in AI, particularly in coding capabilities with a 77% score on Swebench Pro. The model is accessible to select enterprise partners. This strategic approach, alongside a 10% Google investment and notable valuation increase to $30B ARR, positions Anthropic as a key player in the AI landscape, albeit with ongoing debates about accessibility and the balance between innovation and responsible deployment. The industry is also grappling with the implications of AI agents and the future of open-source tools like OpenCL and their potential use in autonomous security research.

Anthropic's Mythos Model: A Cybersecurity Game Changer Stirs Debate on AI Availability and Safety

Anthropic's new Mythos model, while not publicly released due to perceived danger, is being deployed to major companies for cybersecurity testing, demonstrating advanced capabilities in vulnerability detection. This move highlights a growing trend of AI models being restricted to select entities, sparking debate about accessibility and the potential for a "permanent underclass" in AI development. The model's high performance in coding benchmarks and autonomous vulnerability discovery raises both excitement and concern within the AI community.

Redefining Startup Ecosystem Building with Optionality and Intentional Design

Traditional metrics like job creation inadequately assess startup ecosystem success; instead, "optionality" for workforce mobility and skill transference within a focused sector is a more relevant indicator. Effective ecosystem development requires intentional design beyond basic meetups and accelerators, encompassing diverse funding models, policy considerations, and university collaborations to cultivate environments where startups, not just general entrepreneurship, can thrive. South by Southwest (SXSW) serves as a unique "soft landing" program for states and regions, fostering collisions between culture, innovation, and venture capital, thereby enabling strategic multi-state collaborations, particularly in shared economic focus areas like aerospace, energy, clean tech, and AI data centers.

Harnessing AI for Autonomous Software Development at OpenAI

OpenAI is leveraging large language models (LLMs) to achieve highly autonomous software development. Their approach focuses on creating an AI-native environment where agents write, test, and even review code with minimal human intervention. This strategy significantly accelerates development cycles and fundamentally redefines traditional software engineering roles, enabling a small team to manage an extremely large codebase.

Prompt Naming Conventions: "Prompt Injection" vs. "Lethal Trifecta"

The discussion highlights a notable difference in the self-evident nature of technical terms, contrasting "prompt injection" with "lethal trifecta." The core insight revolves around how clearly a term's meaning is conveyed through its name, which significantly impacts its adoption and understanding within a technical community. This distinction is critical for effective communication and education in rapidly evolving fields, particularly in AI/ML safety.

OG Image Update for AI Engineer Summit Site

The AI Engineer Summit website, ai.engineer, requires an update to its Open Graph (OG) image. This is a critical task to ensure proper social media previews and branding, as indicated by a direct request from a user referencing the site.

Swyx Investigates an Unspecified Topic

Swyx, a prominent figure, is currently investigating an undisclosed topic, indicating an ongoing inquiry. The lack of specific details prevents further analysis of the subject matter or potential implications. This suggests a potential future announcement or development worthy of monitoring.

Amazon S3 Files Introduces File System Access to Cloud Object Storage

Amazon S3 Files now offers fully-featured, high-performance file system access, differentiating it as the first and only cloud object storage with this capability. This innovation addresses the historical gap in cloud storage by providing traditional file system functionalities alongside object storage benefits, potentially streamlining data management for developers and enterprises. The announcement directly follows discussions around the lack of open-source cloud storage solutions with similar features, suggesting a timely market entry.

OpenAI’s "Extreme Harness Engineering" Achieves Autonomous Code Generation and Review

OpenAI's "Extreme Harness Engineering" initiative, as discussed in the Latent Space podcast, demonstrates a significant advancement in autonomous software development. This approach, exemplified by projects like Frontier and Symphony, enables the generation and daily processing of massive codebases (1M LOC, 1B tokens/day) with zero human involvement in both code creation and review. The methodology effectively creates "software factories" that operate without direct human intervention, raising implications for future software development paradigms.

Podcast delves into lives of token billionaires

A recent podcast episode explores the personal and financial lives of individuals who have amassed significant wealth through cryptocurrency tokens. The discussion aims to provide insights into the experiences and perspectives of these "token billionaires."

Humor and Engagement in Social Media Content

This content highlights the use of humor, specifically encouraging costumes, and direct engagement with "customers" as a strategy for social media interaction. The intent appears to be fostering a more playful and interactive environment on platforms like X (formerly Twitter). This approach aims to boost user participation and brand likeability through non-traditional methods.

Humor and Absurdity in AI Conference Contexts

The user, Swyx, humorously notes the unusual request for "lobsters" at an AI conference. This highlights a potential disconnect between typical conference expectations and a more whimsical or unexpected element being introduced, likely for comedic effect or a unique experience. The "poor guy" he briefed is probably a conference organizer or event planner.

Swyx Activates "LOBSTER_BOOTSTRAP" and Hints at AI/ML Engineer Booth Build Day

Swyx, a prominent figure, has activated "LOBSTER_BOOTSTRAP" and is engaging in a booth build day for @aidotengineer. The context suggests a public-facing event, potentially a conference or trade show, given the emphasis on a "booth build day."

AI Engineer Singapore Event Announced

An AI Engineer event is scheduled to take place in Singapore. The event is likely a networking or community gathering, given the invitation to "meet them," and prominently features a specific URL for registration or further information: ai.engineer/sg. This suggests a direct call to action for individuals interested in AI engineering in the region.