Mei Z. is a Co-founder and Partner at Alloy Fund who previously co-founded subatomiQ, a technical startup backed by Alchemist, giving her hands-on experience building and fundraising for complex companies. She frequently appears as an angel investor in competitive early-stage rounds alongside well-regarded venture firms and is recognized across multiple VC intelligence platforms as a go-to co-investor for technically ambitious startups.

• A strong focus on compute infrastructure and inference scaling. Frequent highlights of the rapidly growing need for inference and raw computing power are backed up by celebrations of early seed investments in specialized infrastructure startups like Nscale and MatXComputing.
• Deep interest in shifting market dynamics and the difficulty of building defensible moats. While launching a startup is easier than ever due to autonomous tools, establishing a true structural advantage is noted as incredibly difficult. This is illustrated by pointing out that most major industry exits so far, such as Inflection and Adept, have essentially been talent acquisitions of companies with zero revenue rather than traditional public offerings.
• Firm belief in a fast approaching timeline for artificial general intelligence. A standing prediction marks 2027 as the arrival year for this milestone, driven specifically by the rapid advancement of autonomous code agents that are expected to drastically reshape the broader economy.
• Fascination with advanced reasoning and how models generalize knowledge. There is ongoing discussion about inference time reasoning and a specific hope that when models eventually conquer complex benchmarks like international math olympiads, those underlying problem solving skills will translate to entirely new domains rather than being strictly limited to mathematics.
• Attention toward the training data bottleneck and its practical solutions. With the industry essentially exhausting standard internet data sources, generating synthetic data and mass transcribing audio from massive video platforms are highlighted as crucial next steps for developers trying to train the next generation of models.