AJ Asver

Building @ParchaHQ. Prev Product @compound, @brexhq, @coinbase, @google. Alum @ycombinator, @uniofoxford. Dad of twins, in-person maximalist, Coda Yoda.

[AJ Asver @_aj]

At @ParchaHQ we saw coding costs explode so we built our own IDE that does routing at the harness level e.g. Claude Code for planning, Codex for long running execution, Cursor for bugs. Plus custom routing, remote sessions, a browser inspector and voice interface.

Now it's open-source! 👇

[AJ Asver @_aj]

@AmitTandonAI @FactoryAI @harvey @FireworksAI_HQ A fourth one is routing at the harness layer, i.e. between harnesses, depending on the coding task.

[AJ Asver @_aj]

@GergelyOrosz Definitely resonates with what we experienced and what we're seeing from other eng leaders in our network. It inspired us to finally release this our cross-agent IDE too: https://t.co/jC7DgGuR0H

[AJ Asver @_aj]

At @ParchaHQ we saw coding costs explode so we built our own IDE that does routing at the harness level e.g. Claude Code for planning, Codex for long running execution, Cursor for bugs. Plus custom routing, remote sessions, a browser inspector and voice interface.

Now it's open-source! 👇

[AJ Asver @_aj]

@Kyriakos_Pelek @ParchaHQ Pretty straightforward - Check out the code - it’s OSS.

[AJ Asver @_aj]

At @ParchaHQ we saw coding costs explode so we built our own IDE that does routing at the harness level e.g. Claude Code for planning, Codex for long running execution, Cursor for bugs. Plus custom routing, remote sessions, a browser inspector and voice interface.

Now it's open-source! 👇

[AJ Asver @_aj]

@tsirbiladz3 @ParchaHQ Exactly!

[AJ Asver @_aj]

@souvikdeb26 @ParchaHQ Eleven labs - instructions to set up in the Readme

[AJ Asver @_aj]

@akhil_kestur @ParchaHQ We use a meta harness that runs Cerebras and does it based on the user request category. You can change this in the settings and add your own categories

[AJ Asver @_aj]

@jason_haugh @ParchaHQ it uses cerebras to manage the routing and transfer of context via a Flue-based harness

[AJ Asver @_aj]

Here's the link: https://t.co/gBgVYDmBuP

[AJ Asver @_aj]

Here's the link: https://t.co/gBgVYDmBuP

[AJ Asver @_aj]

At @ParchaHQ we saw coding costs explode so we built our own IDE that does routing at the harness level e.g. Claude Code for planning, Codex for long running execution, Cursor for bugs. Plus custom routing, remote sessions, a browser inspector and voice interface.

Now it's open-source! 👇

[AJ Asver @_aj]

Here's the link: https://t.co/gBgVYDmBuP

[AJ Asver @_aj]

At @ParchaHQ we saw coding costs explode so we built our own IDE that does routing at the harness level e.g. Claude Code for planning, Codex for long running execution, Cursor for bugs. Plus custom routing, remote sessions, a browser inspector and voice interface.

Now it's open-source! 👇

[AJ Asver @_aj]

Here's the link: https://t.co/gBgVYDmBuP

[AJ Asver @_aj]

You can check out the Build IDE here: https://t.co/zGiDmubc3J

[AJ Asver @_aj]

I think routing now actually needs to happen at the harness, not model level, because each model is now optimized for their native harness. e.g., Claude Code w/ Opus for planning, then Codex for 80% execution, and Cursor Composer for the last 20%, which is actually 80% of the spend.

[AJ Asver @_aj]

Link to full article: https://t.co/adytpsFuPX

[AJ Asver @_aj]

Awesome to see @gilbert sharing how they use @Grepdotai for episode research on @AcquiredFM in a GQ article this week.

There are so many non-obvious use cases for doing deep research on a business, and Grep is the best at it.

[Miguel Ríos Berríos @MiguelriosEN]

We're subletting space (desks + a private room) in our office in FiDi. Right on the best block in the city — a crosswalk away from @UseCorgi's cafe. DM if interested.

[AJ Asver @_aj]

“On a standalone basis, Starlink is the most profitable satellite business ever built — and arguably the most profitable connectivity business of any kind.”

[AJ Asver @_aj]

Grep’s deep research is very good at analyzing S-1 filings!

[Miguel Ríos Berríos @MiguelriosEN]

$SPCX S-1 analysis by @grepdotai https://t.co/iGh1NOMcyI