Pulkit

Founding Engineer @Vorfluxai

[Santiago @svpino]

This is controversial, but in many cases, humans have become the bottleneck in building software.

You are making everything slower and worse by babysitting agentic coding tools. You aren't better at coding than frontier models, and you are letting your ego get in your own way.

I'm working with @vorfluxai and their platform to build software on autopilot.

"Autopilot" is the key here: their platform is designed to work autonomously, for a long time, with no human intervention.

I like it because they do a lot of the verification I'd have to do myself.

An idea goes in → good software comes out

In my tests, this works well for complex problems and non-trivial tasks.

You don't use this to fix a button's color. You use this to build an entire feature that would take you hours of planning and days of work.

Now, the reason the output is really good is that they have baked in adversarial agents that verify outputs at different levels.

They aren't trying to finish earlier. They are focusing on giving you good software back.

It takes 5 minutes to set up. Connect your repo, give it a task (something big, ambitious), go to sleep, and come back the next day to a solution.

That's it.

By the way, really cool that they built a bunch of ways to notify you when the work is done, and you get a summary video with everything that was done.

Here is the link to their platform: https://t.co/6raqB72VHx

Thanks to the team for partnering with me on this post.

[Vorflux @vorfluxai]

1. Find the bottleneck
2. Thoughtfully apply tokens
3. Test across hundreds of cases
4. Find the next bottleneck + repeat

[Vorflux @vorfluxai]

Autopilot for software engineering, now in your pocket.

Vorflux is on the App Store and Google Play. Your agents plan, build, test, and ship the PR. When they need a call, you take it from your phone.

Stop babysitting your agents.

[Vorflux @vorfluxai]

We're live with this week's episode

https://t.co/zqw0BDm858

[Prasanna S @myprasanna]

Not like this man on the screen. Use @vorfluxai 😊true parallelism out of the box. Assign a linear or a jira ticket and each one gets its own box with your dev env fully running, auth solved, data seeded etc. learns from your previous inputs. Sub agents and configurable harness so you can get reviews from other models or testing done by another model etc. welcome to the cloud.

[Pulkit @St3amWalk3R]

@gaurishcodes @Cali_Burrito No CB???? 😂😂

[Gaurish Baliga @gaurishcodes]

Demoing @vorfluxai
At Codex Community Day Mumbai! 🚀

[Vorflux @vorfluxai]

Opus 5 is now live in Vorflux

[Elon Musk @elonmusk]

@OfficialLoganK Sorry, I got mixed up. Meant to post this one:

[Pulkit @St3amWalk3R]

@myprasanna @ridham_shah_25 @ridham_shah_25 we have reset the balance to 0. I will look into the negative balance issue.

[Prasanna S @myprasanna]

Just wrapped my first X Space with @owengretzinger, Head of Eng @boardyai

Biggest takeaway: AI is making the ticket queue obsolete.

What replaces it is an engineering org built around judgment, customer context, and verification.

My takeaways in the 🧵

[Vorflux @vorfluxai]

We’ll be going live at 10am PST / 1pm EST with @myprasanna and @owengretzinger to talk future of engineering. Tune in

[Vorflux @vorfluxai]

📷 New in Vorflux: Merge Queue

Land PRs safely without the usual "rebase, wait for CI, merge, repeat" dance.Drop a PR into the queue and Vorflux takes it from there.

For each PR it will:
▪️Rebase the PR onto the latest base branch automatically
▪️Resolve merge conflicts on its own when they come up
▪️Run your tests against the rebased commit and only land when it's green
▪️Retry flaky tests automatically so a flake doesn't block your merge
▪️Land PRs in order, one at a time, so main always stays buildable

[Pulkit @St3amWalk3R]

@myprasanna @Nathaliely4 Kimi 3 is live now on @vorfluxai! Happy tokenmaxxing 🚀

[Vorflux @vorfluxai]

🚀New in Vorflux: PR Risk Assessment in Auto Review

Assess the risk of a pull request before approving it.

▪️Each PR gets a risk score from 1–10
▪️Vorflux explains the main risk factors behind the score
▪️If the score is within your configured threshold, Vorflux can approve the PR
▪️If the score exceeds the threshold, Vorflux leaves the risk assessment but does not approve it

How to use
▪️Go to Settings → Pull Requests → Auto Review
▪️Turn on Use risk assessment
▪️Set the maximum risk for approval
▪️Optionally add custom risk assessment criteria
▪️Save the settings

[Prasanna S @myprasanna]

Launch was a 🚀 - thank you all. Kept it short - but @vorfluxai does a LOT more I did NOT talk about. Will talk about 3.

1. Slack/Linear/Mobile app: People just work out of here and never open the Vorflux app. Treat it like an engineer and keep it high level. Supports OpenAI Symphony spec - so you manage higher level.

2. Memory: All sessions can search/explore other sessions, share a /memory/ disk. Common trajectories are dreamt into skills or scripts to save future tokens and improve intel. Self improving.

3. Harness configurator: Backward compatible with Claude/Codex - we auto import your skills and sub agents. Best way to improve intelligence is to recruit sub agents with new models and context windows. Natively support Claude workflow framework so can do massive scale outs. You should try it.

Finally, with our Codex partnership -- you can bring your $200/mo codex plan to Vorflux. You can tokenmaxx like crazy heavily subsidized by OpenAI.

[Ryan Petersen @typesfast]

My favorite launch video I've seen in a while. I don't know quite why but I find it so charming and badass at the same time.

[Esha @eshamanideep]

we've run cursor, devin, and most of the cloud coding agents on real work. @vorfluxai is the only one that stuck with our engineers

why: you compose the harness yourself - fable plans, 5.6 sol builds. sandboxes mirror our actual dev env. and the default harness has a simplifier + reviewer subagent baked in, so the PRs land clean instead of needing a cleanup pass.

also the fastest-moving team we've worked with as a customer!

don't demo it on an easy ticket. give it your hardest one. congrats on the launch @myprasanna

[dCodes @dCodes03]

The team is full of cracked engineers
I tried the product in May
At that time I was able to see many improvement opportunities in the product itself
And today the product looks 100x better than it was at that time
This shows that how much backlog the product it self able to clear
Crazy stuff
Cloud agentic compute is the future….

[Vorflux @vorfluxai]

Introducing, The Open Model Harness 📷

Run full Vorflux sessions on open-weight models.

Select Open Model Harness from the harness dropdown (same place as Opus, GPT, and Fable) and every agent in the session routes to an open model:
⚫ Main, plan, review: DeepSeek V4 Pro
⚫ Explore: DeepSeek V4 Flash
⚫ Design, build, simplify: GLM 5.2
⚫ Debug, testing: Kimi 2.7 Code

You get the exact same Vorflux flow
⚫Full flow included: plans, subagents, PRs, and testing
⚫Excellent price-to-performance ratio
⚫Strong default when you want solid results without premium-model spend

It stays current automatically. As new open weights ship and our internal evals surface better options, we swap in the upgrades. No reconfiguration on your end.