🤖 AI NEWS ROUNDUP — July 06, 2026
1️⃣ ANTHROPIC ENTERS DRUG DESIGN — AI LABS BUILD THEIR OWN PHARMA PIPELINES
Anthropic announced it will develop its own drugs, targeting diseases that traditional big pharma overlooks. The move comes alongside the launch of Claude Science, an AI workbench integrating 60 scientific databases and tools for genomics, proteomics, and chemistry. While skeptics argue the company lacks biotech expertise, researchers like Patrick Kidger counter that building internal drug programs could accelerate AI-for-science capabilities and attract top talent to the field.
🔹
@PatrickKidger
2️⃣ ANTHROPIC'S CLAUDE SCIENCE LAUNCHES WITH AUTONOMOUS RESEARCH CAPABILITIES
Claude Science connects to over 60 scientific databases and can autonomously run single-cell RNA sequencing, CRISPR design, and computational chemistry workflows on HPC clusters. The platform runs on existing Claude models including Opus 4.8. Industry data from BCG shows AI-discovered molecules now achieve 80-90% Phase 1 safety rates versus the historical average of 50%, though Phase 2 efficacy remains at around 40%.
🔹
@m_goes_distance
3️⃣ ANTHROPIC CLAUDE DESIGN TARGETS FIGMA — THE PREDATORY PLAYBOOK RETURNS
Anthropic's Chief Product Officer Mike Krieger resigned from Figma's board on April 14, 2026. Just three days later, Anthropic launched Claude Design, directly competing with Figma's core product. The pattern mirrors similar moves by Big Tech companies entering markets where their executives held board positions, raising questions about competitive dynamics in the design tool space.
🔹
@MilkRoadAI
4️⃣ CEREBRAS CEO: DATA CENTERS, NOT CHIPS, ARE THE AI GROWTH BOTTLENECK
On their first earnings call as a public company, Cerebras CEO Andrew Feldman identified data center availability as the primary constraint on AI infrastructure expansion, surpassing chip supply and customer demand. The insight highlights a shifting bottleneck in the AI buildout — compute hardware is scaling rapidly, but finding facilities with sufficient power and cooling capacity is becoming the limiting factor.
🔹
@BlackPantherCap
5️⃣ TENCENT RELEASES HY3 — 295B MoE MODEL WITH ONLY 21B ACTIVE PARAMETERS
Tencent dropped Hy3, a 295-billion parameter Mixture-of-Experts model where only 21 billion parameters activate during inference, plus a 3.8B MTP layer. In blind tests with 270 domain experts, Hy3 scored 2.67/4 on benchmarks, outperforming GLM-5.1 at 2.51/4. The model is fully open-weight and can run on 4× RTX Pro 6000 workstations, making frontier-class performance accessible to smaller teams.
🔹
@testingcatalog
6️⃣ HY3 BENCHMARKS REVEAL COMPETITIVENESS WITH FRONTIER MODELS
Detailed benchmark analysis shows Hy3 nearly matches GPT-5.5 and Claude Opus 4.8 on BrowseComp, and beats GPT-5.5 on Frontierscience-Olympiad. Tencent trained the model using feedback from over 50 internal product teams, optimizing for execution, interaction, and real workflow performance rather than synthetic benchmarks alone.
🔹
@yuhasbeentaken
7️⃣ AI WEEKLY RECAP — JUNE 29 TO JULY 5 SEES MAJOR MODEL ROLLOUTS
The week from June 29 to July 5 brought a wave of releases: limited rollout of GPT-5.6 with safety-focused updates, Anthropic's Claude Science platform launch, Tencent's Hy3 open-weight model, and continued export control developments affecting Anthropic's most advanced models. The pace of releases across both Chinese and US labs underscores an accelerating competitive cycle in AI development.
🔹
@shushant_l
💭 The AI industry is entering a phase where toolmakers are becoming domain players — Anthropic building drugs, Claude Design challenging Figma, Chinese labs releasing models that rival the frontier. The separation between "AI infrastructure" and "AI applications" is blurring fast, and the companies that can bridge both are likely to define the next cycle.
Which of these developments do you think will have the biggest long-term impact — AI in drug discovery, open-source models catching up to frontier, or data center bottlenecks shaping AI growth? 👇
#AI #Anthropic #Tencent #Hy3 #ClaudeScience #AIDrugDesign #Cerebras #OpenSourceAI