Apple Took Years to Catch Up. Kilo Code Took 6 Weeks--and It's Coming for Lovable, Cursor, Replit

Artificial General Intelligence: The Future is Closer Than We Think

Major Developments in AI Funding and Safety Concerns

• This week, two leading figures in artificial general intelligence (AGI) discussed the rapid advancement of AGI and its potential impact on entry-level white-collar jobs.

• XAI announced a $20 billion Series E funding round, bringing its valuation to approximately $230 billion, positioning it alongside OpenAI and Anthropic.

• Despite facing safety issues with Grock generating inappropriate deep fakes, XAI secured a Department of Defense deal, indicating strong investor confidence in long-term AI value.

• Investors appear willing to overlook initial safety concerns as they believe AI will mature over time; this reflects a long-term perspective on the technology's potential.

Insights from Davos: Perspectives on AGI

• At Davos, Daario Amade from Anthropic predicted that AGI could emerge by 2026 or 2027 due to AI's ability to write its own code.

• Deise Hasabis from Google DeepMind offered a more conservative estimate of a 50% chance for AGI by the end of the decade, emphasizing challenges in job automation.

• Hasabis argued that while AI can enhance productivity, it may not fully replace human roles due to the unique skills humans bring to their jobs.

• Current employment impacts are mixed; while junior positions may be harder to secure, overall layoffs do not seem significantly influenced by AI advancements.

Challenges Facing Current AI Models

• Key areas needing improvement in current AI models include memory retention, continuous learning capabilities, and long-term reasoning abilities.

• Hasabis believes these challenges can be addressed but suggests that significant breakthroughs may take longer than initially anticipated.

Apple and Google's Strategic Collaboration

• Apple has partnered with Google for future foundation models based on Google's Gemini technology—a significant setback for OpenAI.

• The collaboration reportedly costs Apple around $1 billion annually and involves developing a custom 1.2 trillion parameter model tailored specifically for Apple's needs.

Overview of Recent Developments in AI and App Development

Transition to Gemini OS

• Discussion on the potential shift from a secondary operating system (OS) to Gemini being the default OS across platforms, including Android and iOS. This change increases pressure on key figures like Sam Alman and Johnny IV to innovate with a third device for OpenAI distribution.

Introduction of Engram by DeepSeek

• On January 12th, DeepSeek published a paper introducing "Engram," addressing the limitations of transformers in knowledge lookup capabilities. The current models require extensive attention layers for simple tasks, which is inefficient.

• Engram proposes a solution by utilizing short sequences (2-3 tokens) and hash functions to access a large embedding table. This method filters patterns against current context using a gating mechanism.

• The efficiency of Engram allows significant performance improvements without excessive token usage, marking it as an important breakthrough in providing factual memory within models.

Kilo Code's New App Builder

• Kilo Code, founded by GitLab co-founder Sid and CEO Scott, launched its app builder after raising $8 million in seed funding. The company aims to disrupt existing tools like Lovable by targeting engineers rather than non-technical users.

• The app builder differentiates itself by focusing on engineering needs, positioning itself alongside VS Code as an open-source tool that caters specifically to developers' requirements.

Market Dynamics in AI Coding Tools

• As the coding market matures, there is a shift from novelty ("AI can write code") to evaluating which differentiated tools best fit workflows. Kilo's strategy emphasizes reliability and integration over isolated solutions.

• Noteworthy speed in development: five engineers completed an internal demo within three days; public launch occurred six weeks later. Their roadmap suggests rapid iteration based on customer feedback could lead to quick improvements.

Competitive Landscape Analysis

• Questions arise about whether there is space for another player like Kilo Code between established tools such as Lovable and Cursor. Cursor has gained popularity among engineers while Kilo seeks to carve out its niche.

• Anticipation exists regarding continued innovation within the AI agent coding market as it becomes clear that enhancing tools for engineers will be crucial moving forward into 2026.