PWV Portfolio Feed

The Elicit team (Xiangxu Kong and Prem Mohanty) is on the road at PEGS Boston 2026 next week (May 11-15). 

Be sure to attend the presentation by Xiangxu Kong to learn more about AI for literature search and evidence synthesis.

Visit us at Booth # 35 for demos and discussions. See you there!

tell that to the people driving by my office and apartment! i hear it at least a few times a week

new york city is one of the least cringey places in the world, with the exception of people's obsession with the song 'empire state of mind'

(The Death And Life of Great American Cities)

Also I hate to say this because it's *always* my dad's book criticism and I often disagree—but it feels like the book's twice as long as it needs to be

I'm reading Jane Jacobs for the first time, and I'm interested but tbh it's a bit of a slog. Part of the problem is just that the neighborhood examples are over half a century old, so maybe I need a condensed, updated version

With Unikraft, Prisma Postgres reduced its PostgreSQL image from 280MB to 61MB.

These smaller images run as unikernels on bare metal.

That gives each database strong VM-level isolation, while allowing a single machine to host thousands of Prisma Postgres instances.

Breakdown 👇

post did well on hacker news! i think it had the critical ingredients of:

- not using robots to write
- non-attachment to "success metrics"
- pointing at the yawning chasm between how companies brag about their good code and their engineers whisper about the trashfire of tech debt

It's easy to introduce slow queries into your workload. 

Tracking them down should be easy too.

PlanetScale Insights shows you which queries are slow and why. Combined with our MCP server, you can automate fixing them.

We also evaluated our newest screening and extraction models across 108 Cochrane reviews to achieve:
- 95% recall with semantic search alone
- 97% sensitivity & 93% specificity on abstract screening
- 99.5% sensitivity & 70% specificity on full-text screening
- 96% accuracy in data extraction

Major upgrades include:
- Search, screen, extract up to 40,000 papers in real-time
- Dual review
- Full-text screening
- Decision logs & exclusion reasons
- Multiple search strategies across multiple databases
- 100% reproducibility for keyword search
- PRISMA diagrams
- Report templates
- The API