SunBrief#67: OpenAI and Anthropic Launches New Models Same Day

Welcome to the SunBrief

Today in SunBrief 🌞

• Leadership Can’t Be Automated

• Anthropic Launches Claude Opus 4.6 while OpenAI Strikes Back with GPT-5.3-Codex

• Conduent Data Breach Expands Dramatically, Affecting Tens of Millions Across U.S.

• Stock Updates

• AI Tackles Labor Shortage in Rare Disease Treatment, Boosting Biotech Innovation

• AI Highlights of the Week

• Too Important to Miss

Leadership Can’t Be Automated

AI can help you move faster, but real leadership still requires human judgment.

The free resource 5 Traits AI Can’t Replace explains the traits leaders must protect in an AI-driven world and why BELAY Executive Assistants are built to support them.

Anthropic Launches Claude Opus 4.6 while OpenAI Strikes Back with GPT-5.3-Codex

On February 5, 2026, Anthropic released Claude Opus 4.6, its most advanced model yet just 27 minutes before OpenAI launched GPT-5.3-Codex, raising the bar in the AI rivalry.

Claude Opus 4.6 Key Upgrades

• Agent Teams: Multiple Claude agents can now collaborate autonomously on tasks like code reviews.

• 1M-token context window (beta): Stronger long-context memory and retrieval.

• Top scores on coding (Terminal-Bench 2.0), reasoning (Humanity’s Last Exam), and knowledge work (GDPval-AA).

• New tools: Claude in Excel (smarter unstructured data handling) and PowerPoint (research preview).

• Safety-first: Industry-leading low misalignment and refusal rates.

Available now via claude.ai, API, and major cloud platforms.

GPT-5.3-Codex Key Upgrades

• State-of-the-art coding across SWE-Bench Pro and Terminal-Bench 2.0.

• Full-spectrum agent: Builds software, writes decks, analyzes data, and responds in real time.

• Built to build itself: Used early versions to debug and optimize its own training.

• Cybersecurity focus: First “High Capability” AI for cyber tasks, backed by $10M in API credits for defense use.

Live now in ChatGPT (paid), CLI, IDE, and Codex app. API access coming soon.

Comparison Between Two: 

• Best use-case : Opus 4.6 = deep research + massive document reasoning; GPT-5.3-Codex = large codebases + security reviews.

• Context : Opus 4.6 offers up to 1M tokens (beta) (with high max output); GPT-5.3-Codex sits around 400k context.

• Reasoning/knowledge work : Opus 4.6 leads on reasoning + retrieval benchmarks (e.g., GDPval-AA, ARC-AGI-2, BrowseComp).

• Agentic coding : GPT-5.3-Codex leads in terminal/OS-style coding tasks (e.g., Terminal-Bench 2.0, OSWorld-Verified).

• Efficiency + iteration : GPT-5.3-Codex emphasizes speed/token efficiency, while both labs highlight models accelerating their own build cycles (with safety/reward design becoming more important).

Conduent Data Breach Expands Dramatically, Affecting Tens of Millions Across U.S.

Stolen Data Includes Social Security Numbers, Medical, and Insurance Information

A January 2025 ransomware attack on government tech giant Conduent has ballooned in scale, now impacting at least 25.9 million people across Texas and Oregon, with millions more affected in other states.

Key Points:

• Massive Exposure: Texas alone accounts for 15.4 million impacted individuals, while Oregon adds 10.5 million. Delaware, Massachusetts, and New Hampshire also reported affected residents.

• Sensitive Data Compromised: The breach includes names, Social Security numbers, health records, and insurance data, some of the most sensitive personal identifiers.

• Scope Still Unknown: Conduent, which supports services for over 100 million Americans, has not confirmed a total number of victims and continues to send notifications.

• Delayed Disclosure: The company reported the attack in April 2025, months after systems were knocked offline, impacting U.S. government services.

• Attack Attribution: The Safeway ransomware group claimed responsibility, alleging theft of 8TB of data.

Why It Matters:
With operations spanning dozens of government programs and corporate services, Conduent’s breach could rank among the largest in U.S. history. The lack of transparency and delay in notification have raised urgent concerns about cybersecurity practices in public-sector tech vendors.

Stock Updates

AI Tackles Labor Shortage in Rare Disease Treatment, Boosting Biotech Innovation

Insilico and GenEditBio use AI to accelerate drug discovery, gene editing, and clinical trial design

Modern biotech companies like Insilico Medicine and GenEditBio are harnessing AI to overcome a critical challenge: the shortage of skilled labor in treating thousands of rare diseases that remain without cures.

Key Points:

• Pharmaceutical Superintelligence: Insilico’s “MMAI Gym” trains generalist AI models to tackle multiple drug discovery tasks with high precision, from identifying disease targets to repurposing existing drugs.

• Automated Discovery: Their AI-driven platform analyzes biological, chemical, and clinical data to generate hypotheses and design new molecules cutting time and cost dramatically.

• Gene Editing Revolution: GenEditBio develops engineered protein delivery vehicles (ePDVs) that use AI to deliver gene-editing tools inside the body with tissue-specific precision.

• FDA Green Light: GenEditBio recently gained FDA approval to begin trials on a CRISPR therapy for corneal dystrophy.

• Data as Bottleneck: Both companies note a persistent challenge: access to unbiased, high-quality biological data. They are investing in automated labs and large in vivo testing pipelines to generate new datasets.

Why It Matters:
AI is becoming the force multiplier biotech needs, not just to accelerate discovery, but to democratize treatments for neglected rare diseases. The fusion of lab automation, gene editing, and AI may redefine the pace and scale of medical innovation in the decade ahead.

AI Highlights of the Week

• Anthropic Introduces Claude Opus 4.6 with 1M Token Context

  Anthropic has unveiled Claude Opus 4.6, its smartest model yet, delivering stronger coding, reasoning, and agentic task performance.

  The model adds a 1M token context window (beta) and state-of-the-art benchmark scores, powering advanced knowledge work, research, and autonomous workflows across Claude and its API.

• Apple Scales Back AI-Powered Health Coach Plans

  Apple has reportedly scaled back Project Mulberry, its planned AI-driven health coach, after recent leadership changes.

  Some AI health features may still ship gradually, including chat-based guidance and enhanced Siri health queries in future iOS releases.

• Amazon Stock Drops 10% After Q4 Profit Miss, $200B Capex Plan

  Amazon shares fell over 10% after the company missed Q4 profit expectations, despite posting record revenue and strong AWS growth.

  CEO Andy Jassy signaled $200B in 2026 capital spending, driven largely by AI and cloud infrastructure, raising investor concerns over rising costs.

• Claude Will Remain an Ad-Free Space to Think

  Anthropic says Claude will stay ad-free, arguing that ads don’t belong in AI conversations built for focus and deep thinking.

  The company will rely on subscriptions and enterprise revenue, keeping Claude aligned solely with user interests, not advertisers.

Too Important to Miss

Cerebras Systems Raises $1 Billion to Push the Limits of AI Computing

OpenAI Launches Frontier to Turn AI Agents Into Enterprise Coworkers

Perplexity Introduces Model Council for Multi-Model AI Answers

Amazon Brings OpenAI Technology Into Alexa+

Last Week’s Poll Result

• Do you think AI-generated game worlds like Project Genie are the future of game creation?

  Maybe, but traditional game development will still matter → 45.00%

  Yes, anyone will be able to build worlds with AI → 40.00%

  No, AI worlds feel too shallow → 15.00%

• Do you agree with Anthropic’s stance against AI use in autonomous weapons targeting?

  Yes, human oversight must be mandatory → 66.67%

  No, AI should be used wherever it’s effective → 23.81%

  Maybe, but AI can assist under strict rules → 9.52%

• Do you think AI will eventually outperform human experts in interpreting genetic variants?

  Maybe, as a decision-support tool → 50.00%

  Yes, at large scale it already is → 37.50%

  No, human expertise will remain essential → 12.50%

Feedback

Know someone who may be interested?