Tech news picked on 2026-04-08
How a 26-Person Team Threatens Big Tech AI
- Miami-based startup Arcee AI, consisting of 26 people, spent 6 months and $20 million to develop an open-source model with 400 billion parameters.
- They used 2,048 Nvidia Blackwell B300 GPUs and targeted a niche focusing on small enterprise models and deployment within VPCs, rather than competing in the general LLM arena.
- Even small teams can turn their choice of technology into a competitive advantage by first defining their target customers and deployment environments.
- However, creating the model and getting customers to pay for its difference are separate challenges.
From a 6 Billion Won Loss to a 1.8 Billion Won Operating Profit: Enrise's Turnaround
- Enrise, which operates the home training app Quat and dating app Wiffy, posted an operating profit of 1.82 billion won in 2025.
- After a 6 billion won loss in 2022, it achieved two consecutive years of profit.
- Wiffy's earnings were used to grow Quat, and within a year of its launch, Quat accounted for 60% of sales, securing profitability.
- It was less about running multiple businesses and more about the timing aligning well.
- However, it is still unproven whether the same subscription model will work in Japan.
Eclipse's $1.3 Billion Fund Marks the Dawn of the Physical AI Era
- Venture capital firm Eclipse has created a new $1.3 billion fund dedicated to Physical AI.
- They plan not just to invest in existing startups but to incubate by building and growing companies directly.
- This signals a shift of capital toward AI operating in the physical world, like robots and edge devices.
- However, large funds do not buy vague hardware ideas.
- It is advised to narrow down to one task to prove at the first customer site and the cost items to verify first.