PWV Portfolio Feed

Capability 5: Knowledge that compounds over time

Improve each decision with Projects that keeps sources, context, instructions, and outputs connected in a single knowledge base.

Capability 4: Transparency by default

Trust and defend every decision with transparent methods trail and sentence-level citations.

Capability 3: Decision-ready outputs

Move from research to decisions in a single workflow by turning evidence into fully cited reports, tables, analyses, visualizations, slides, and documents.

Capability 2: Diverse data sources

Make decisions based on the full picture by synthesizing evidence across scientific literature, clinical trials, patents, regulatory and commercial sources, the web, and your own internal data.

We built the Research Agent around five key capabilities:

Capability 1: A harness with reasoning principles 

Reason with greater confidence with built-in models trained on BioDecisionBench, a benchmark designed to evaluate reasoning quality and handle uncertainty around decisions in pharma.

An agent fumbles a task against your service and the instinct is to blame the model. Usually the problem is yours: a slow endpoint, or a tool description that invites the wrong call. At RenderATL, Sean Roberts will score a real project live with AXIS, the open-source agent experience framework.

Jack Herrington thinks TanStack should be your entire web platform, and he's bringing that argument to RenderATL. He'll show TanStack Start for full SSR, TanStack AI for realtime chat and code mode, and TanStack DB for realtime data with any database. All of it is open source and deploys anywhere.

At RenderATL, Nahrin Jalal will build and deploy an app from scratch in 30 minutes using Agent Runners. Her session covers the parts people actually get stuck on: effective prompts, agent context, and connecting the right tools. Everyone walks out with something live in production.

true story, ran a race tonight with the team name “Modena pigeons” and won best team name

Nowhere near as easily, and with several caveats 

With previous projects we’ve found the native tls crate to be more effective for this need

We've migrated our TLS backend from rustls to native-tls, meaning Atuin now uses your operating system's certificate store. 

This fixes atuin for self-hosted servers behind self-signed or internal CA certificates, and on corporate networks with TLS-intercepting proxies

Previously, atuin update lived in a separate atuin-update binary installed alongside Atuin. 

It's now part of the main binary, and lays the groundwork for automatic updates in a future release.

If you're running an old config, switch to this with

atuin config set daemon.enabled true
atuin config set daemon.autostart true
atuin config set search_mode "daemon-fuzzy"

We've replaced the search with frizbee, a SIMD-accelerated fuzzy matcher that benchmarks meaningfully faster than our previous engine (nucleo) for our workload. 

Results are now ranked primarily by match quality, with frecency (how often + how recently you ran a command) breaking ties

atuin v18.19 is out now! 🐢✨

- search is more than 5.2x faster
- self-signed certs and system cert stores just work
- atuin update is a built-in command

Get the latest with `atuin update`!