NEW LAUNCHES

The latest features, products & partnerships in AI

• AI fintech company Jump secures $20m in Series A funding

• Apple’s ELEGNT AI aims to make home robots feel like companions

• Microsoft Edge to start blocking scareware with new AI feature

• Craft app’s latest AI upgrade may change how you use your iPhone forever

• HR tech company Phenom expands AI offerings with acquisition of EDGE

• Meta unveils AI tools to boost ad campaign effectiveness

• Honeywell partners with Microsoft to deploy AI across global operations

• Google Photos adds crucial AI safeguard to enhance user privacy

• Snap targets small businesses with new AI advertising tool

• AI monitors risks to undersea infrastructure in new Windward tool

• Mistral’s AI chatbot ‘Le Chat’ now available as iOS app

MONEY FLOWS

The latest AI deals, rounds & projections

• Venture Capital 3.0: What founders need to know about investors in the AI era

• Ilya Sutskever’s startup Safe Superintelligence may raise funding at $20B valuation

• General Catalyst and MGX in talks to join Anthropic’s massive funding round

• Private equity consortium buys Japanese chipmaker FICT for $660M

IMPLEMENTATION

Announcements, strategies & case studies

• Nearly half of Americans now prefer AI over humans to do their taxes

• A quick guide to implementing AI at your organization

• AI breast cancer detection trial begins in the UK

• How Dialpad achieved $300M ARR by building its own in-house AI

• AI transforms college, pro sports entertainment ROI with 220M data points

• AI app detects skin cancer with 99.8% accuracy

• How smart teams should update technical interviews for software recruiting in the AI era

AI MODELS

Deployment, research, training & infrastructure

• How and when AI models learn to deceive their creators

• BOLT: The new technique that enables AI models to reason through complex problems

• How chain-of-thought prompting hinders performance of reasoning LLMs

• Hugging Face’s new open-source AI model lets robots follow verbal commands

• Ark series is a new ‘large vision language model’ purpose-built for finance and accounting

• Epoch overhauls its AI Benchmarking Hub to improve AI model evaluation

AI AGENTS

Launches, research & more from the AI Agent Report Newsletter

• Researchers claim they’ve created open-source version of OpenAI’s newest AI agent — in only 24 hours

• It’s time to build apps and security protocols for a new type of user: Autonomous agents

• Why 2025 will be the year of the AI App (and not foundation models)

• DeepSeek security vulnerabilities offer glimpse of true problems lurking in the agentic age

• Project Padawan: GitHub Copilot unveils fully autonomous coding agent

• AI agents, decentralized intelligence and the future of collaboration

• The argument against fully autonomous AI agents

IN OTHER NEWS

Compelling stories beyond the usual categories

• DoD awards BigBear.ai contract to identify geopolitical trends from foreign news media

• With AI search threatening click-through rates, connected television may be advertising’s next frontier

• Oscars may mandate AI disclosure for films after recent controversies

• Oxford researchers use AI to decode ancient Greek scroll

• Christie’s to hold its first ever auction dedicated exclusively to AI-generated art

• Google admits Gemini AI demo was staged in Super Bowl ad

• AI accelerates lab-grown meat development with implications for animal welfare, environment

What’s happening in AI right now

Why open source AI is challenging tech giants’ models

The AI world's fiercest battle isn't between competing tech giants - it's between proprietary and open source approaches. This week's developments illustrate how quickly open source teams can replicate supposedly unique AI capabilities, raising profound questions about sustainable competitive advantages in AI.

Rapid replication in action

When OpenAI launched its Deep Research feature, it probably didn't expect Hugging Face to create an open source version within 24 hours. The replicated version achieved 55.15% accuracy compared to OpenAI's 67.36% - remarkable given the timescale. This follows a pattern we've seen repeatedly, where proprietary AI features face swift open source competition.

The trend extends beyond research tools. Hugging Face and Physical Intelligence just released Pi0, an open source foundation model enabling robots to understand verbal commands across multiple platforms. Meanwhile, Chinese AI firm DeepSeek has demonstrated impressive results with fewer resources than industry standards, though allegations about training data misuse have complicated their story.

The speed premium

Speed is becoming a crucial differentiator. Cerebras Systems and Mistral AI's partnership has produced record-breaking response speeds of 1,000 words per second for their Le Chat application. This focus on computational efficiency rather than just raw model size suggests companies are finding new ways to compete.

Specialized solutions emerge

While general-purpose AI models grab headlines, specialized applications are gaining traction. The Ark series of models, purpose-built for financial document processing, demonstrates 2.5x faster processing than human benchmarks. BOLT, a new method for enabling AI reasoning through complex problems, shows how targeted solutions can advance specific capabilities without massive resource requirements.

The moat question

These developments challenge traditional assumptions about competitive moats in AI. If sophisticated capabilities can be replicated so quickly, what sustains long-term advantage? The answer may lie in systems of intelligence that combine multiple specialized components rather than single breakthrough models.

Meta's brain-typing AI research offers an interesting parallel - while achieving 80% accuracy in translating brain activity to text, its real value may be in the insights gained about human cognition rather than the specific implementation.

Looking ahead

As we move into 2025, the industry appears to be shifting from a focus on foundation models to practical applications and integration. The democratization of AI capabilities through open source alternatives suggests success will increasingly depend on how companies implement and combine AI tools rather than purely on model ownership.

We publish daily research, playbooks, and deep industry data breakdowns. Learn More Here