
Community Manager @Studio1HQ || Building Communities and solutions || Qwen Ambassador || Kolkata ||
Kolkata, India
Joined April 2019
- Tweets1.4K
- Following410
- Followers133
- Likes3.9K
Indrajit Ari retweeted
Claude Fable 5.2 is making one thing very clear: the model is only half the system. The harness decides how well everything works.
Anthropic's prompting playbook for agent harnesses has been turned into a single prompt for Claude Code.
It asks about your tech stack first, then builds the entire agent workflow:
→ Orchestrator: breaks the goal into tasks, each with its own completion check.
→ Builders: one agent per task, each with a clear brief and low effort.
→ Subagents: handle parallel work when it makes sense, without spawning agents unnecessarily.
→ verify(): an independent reviewer reads the diff and runs tests.
→ Evals: score every change and track token usage per change.
The prompt also builds in some important rules:
→ Use high effort by default, then rerun evals at medium and xhigh.
→ Keep context append-only and use compaction to manage it.
→ Memory notes must include scope and reasoning.
→ Reads, tests, and safe retries need no approval.
→ Risky actions always require permission.
It even handles common failure cases like timeouts, empty results, tool errors, and files that must not be touched.
No arbitrary token budgets. No blindly delegating every task.
On the first run, it creates a 5-line plan, builds the solution, and reports what worked, what failed, and what's next.
The bigger shift is that we're moving beyond writing better prompts. We're designing the system around the agent.
Worth trying if you're building serious workflows with Claude Code.
Indrajit Ari retweeted
Grateful to have @Lvlupvc as the first investor in our pre-seed round. We’re filling the rest of it with investors who back infrastructure and AI labs early, before the revenue chart. DMs are open.
Welcome aboard, @Lvlupvc, our first investor.
Money moves across systems that don’t agree with each other. Reconciliation is how you prove every ledger tells the same story, and it’s still mostly manual. We’re building the infrastructure to change that, starting with cross-border payments and expanding across industries.
Indrajit Ari retweeted
We're cooking some cool stuff @Studio1HQ
Stay Tuned. 😎
Btw, we're Hiring: studio1hq.com/careers
Joined @Studio1HQ as Community Manager!
More crazy events, bigger plans, and lots of fun coming soon 👀🔥
Joined @Studio1HQ as Community Manager!
More crazy events, bigger plans, and lots of fun coming soon 👀🔥
Indrajit Ari retweeted
We're hiring a Growth guy at @Studio1HQ
please find the JD here: studio1hq.com/careers/growth…
My MacBook suddenly shut down and restarted multiple times, and now my heart is racing rn💀
I genuinely feel like I’m about to have a heart attack.
1
78
Indrajit Ari retweeted
Arnab Goswami has spent years teaching India what a fearless man he supposedly is.
A man who screams at people, humiliates guests & mistakes being an asshole for journalism.
There is a huge difference between being obnoxious in front of your cameras and having the courage to face someone outside your studio.
One requires dubious funding, many microphones, cameras & curated actors who’d submit to this psychopaths ideas.
The other requires courage.
Reuploading this to remind him how
he will get humbled when this government & his perks run out.
Arnab you are & will always remain to be a Coward.

















