Lovepreet Singh

If you’ve shipped an agent, you probably know this loop: → Check logs → Rewrite prompt → Deploy → Hope it works Repeat every week. That’s why this caught my attention: @FutureAGI_ just open-sourced their entire platform Not an SDK. Not a “lite version”. → Full UI → Backend → Simulation engine → Eval + optimization loop All in one repo (Apache 2.0): github.com/future-agi/fut…

@Fardinsk_ A lot of apps or daily useful apps we can build in cli using python scripts and bash. Lot of tokens and mental bandwidth wastage adopting to Electron, react, swift etc

@SinghDevHub why bhai?, could you elaborate

Bash scripting is still one of the most underrated skill in 2026

It is a creator economy. Everyone is building products but creators are earning more than some founders

guys have anyone of you setuped complete e2e claude ?? ex:- - you can check on phone what is happening in the system by claude - claude is running tests, using tool calls to browse etc, using mcps whenever possible - claude is using browser access to run tests or app access to run tests end to end - different claude running on diff worktrees What all GUIs, tools, editors, apps etc you have used to pull this off ?

Posts like this are why we open sourced in the first place. Someone running it, breaking it down honestly, and writing what they actually saw. Means a lot Swapna @swapnakpanda

If you told me two years ago that the hardest part of building self-improving AI would be writing a post about it - i would have laughed. and yet here i am. draft number ‘i-lost-count’ @FutureAGI_ is live on Product Hunt. open source. full platform. hundreds of teams already build self-improving agents on it. the whole lifecycle in one repo. This post isn't perfect. but the platform is pretty damn close. Would appreciate all the support we can get today- 📷 producthunt.com/products/futur…

@sakshamred 😅😂

Peak engineering

@shydev69 😅 lets goo

@SinghDevHub banger 10000 likes #soon

Introducing MacMsi Pro my mac screen got damaged and got to know it will cost 90k to replace it so i bought external portable monitor worth 12k & using it now as my side machine.

@filiksyos Someday 🤞

@SinghDevHub thanks lovepreet.. you should do open source project

gitreverse hit 800 github stars sweeeeet

github.com/future-agi/fut… Most teams building AI agents are stuck in this loop: logs → tweak prompt → redeploy → repeat @FutureAGI_ just open-sourced a full stack to break that 👇 • Full platform (UI + backend + simulation + evals) — not just SDK • Built around a closed feedback loop (fail → fix → validate → redeploy) • Auto-generated adversarial, multi-turn simulations • Built-in evals: hallucination, grounding, tool use, safety, PII • Real-time guardrails (jailbreaks, prompt injection, toxicity) • OpenTelemetry-native tracing • Works with LangChain, LlamaIndex • Self-hostable (keep your data) Big idea: Stop babysitting agents → start systematically improving them This is the infra most teams are duct-taping together manually. Now it’s one system.

If you’ve shipped an agent, you probably know this loop: → Check logs → Rewrite prompt → Deploy → Hope it works Repeat every week. That’s why this caught my attention: @FutureAGI_ just open-sourced their entire platform Not an SDK. Not a “lite version”. → Full UI → Backend → Simulation engine → Eval + optimization loop All in one repo (Apache 2.0): github.com/future-agi/fut…

But the real idea is the optimization loop: When something fails: Simulate root cause Detect issue Generate candidate fix Validate on real traffic patterns Check regressions Re-deploy That loop is what most tools are missing. To be clear — this isn’t “magic self-healing AI” You still need good evals and oversight and you get option to build your own evals on the platform. It’s the closest I’ve seen to systematically improving agents instead of babysitting them Other things worth noting: OpenTelemetry-native tracing Works with LangChain, LlamaIndex, etc. 100+ model providers via gateway Self-hostable (free & you keep your data) If you’re building agents in production: This is basically the infra layer many teams are already trying to build… manually. Now it exists as a single system. Try it yourself: Cloud: shorturl.at/9rvCX (also has a free tier) Repo: github.com/future-agi/fut… #futureagi #futureagi-oss

The most interesting part: simulation Instead of writing test cases manually… It generates adversarial, multi-turn conversations based on your agent’s behavior. The kind of edge cases users hit that you’d never think to test. Then comes evaluation: Hallucination Groundedness Tool usage PII / safety Custom metrics Runs across text, image, audio. And it’s fast enough to run in production. Guardrails are built-in (not bolted on later): → Jailbreak detection → Prompt injection → PII leaks → Toxicity All running in real time on every interaction.

Excited to see Traycer in action! We'll be watching and ready to help if you need any support. Happy building! 🙌

@niveditjain cfbr

I am hiring a Founding Engineer