Krish Chelikavada (@sachelik) profile picture

Building Jina - AI QA Engineer, @0xpass_io | @Stanford, @alliancedao, @zfellows

San Francisco
Joined August 2017
Photo by Krish Chelikavada (@sachelik)Photo by Krish Chelikavada (@sachelik)Photo by Krish Chelikavada (@sachelik)Photo by Krish Chelikavada (@sachelik)Photo by Krish Chelikavada (@sachelik)Photo by Krish Chelikavada (@sachelik)Photo by Krish Chelikavada (@sachelik)Photo by Krish Chelikavada (@sachelik)Photo by Krish Chelikavada (@sachelik)Photo by Krish Chelikavada (@sachelik)
Companies worth >$50B use Jina, our AI QA Engineer, to catch critical issues. Jina simulates production scenarios for every PR and executes code to catch runtime issues. Our new harness takes this deeper.
1
3
1
15
761
Our harness has three roles: - Explorers trace code and downstream effects - Investigators run code and test scenarios - The orchestrator learns from both and adds scenarios Exploration and investigation can happen in parallel and the orchestrate iterates on their results
1
2
86
As results arrive, the orchestrator adds scenarios or digs deeper into unanswered questions. Each follow-up builds on prior evidence. Jina goes broad and deep, adapting as it investigates. Want access? Reach out or sign up - usejina.com
2
56
landed our biggest customer yet 👀 - $60B co Our harness + context advantage has been compounding and we’ve been surfacing some really deep issues from our prod scenario simulations DM me if you care about shipping bug free code
1
1
6
812
Krish Chelikavada retweeted
You can buy Google You can't buy Waymo Daybreak splits a stock into its individual businesses We let you back the part you actually believe in
Daybreak (@daybreak_fi — ALL15) splits tokenized stocks into their underlying business lines, so you can trade Starlink inside SpaceX or Youtube inside Alphabet. Companies are portfolios of businesses. Now you can trade them that way. Congrats on the launch @ademplabbers and @srhhadron
63
27
12
185
37,674
if you're not using jina you're missing out ...
6
242
Krish Chelikavada retweeted
⏰ 3 days left for the Jina campaign! An AI teammate for your PRs, built for developers. It flags what static review misses. 300 Credits for connecting GitHub & X 6.5K Credits pool for the best selected feedback metopia.xyz/campaign/jina
Jina Campaign is live on Metopia! Meet Jina by @sachelik, an autonomous AI QA engineer for pull requests. Test it, share your experience, and earn Jina Credits! 1️⃣ Link and verify your GitHub & X accounts to claim 300 Jina Credits. 2️⃣ Submit your feedback for prizes: 🥇2,000 Credits × 1 🥈1,000 Credits × 2 🥉500 Credits × 5 *Winners will be selected based on the depth and actionability of their feedback. ⏰ Ends: Aug 31, 2026 Join here: metopia.xyz/campaign/jina
1
7
919
congrats on the launch @morganlai
Prediction markets are an information game. Most people play it blind. Ares Mobile Beta is live: social trading for prediction markets, where every opinion has real money behind it. Powered by @Polymarket First trade is free. Drop your position card below → 25 winners get $25 announced on 8/27.
1
3
504
Krish Chelikavada retweeted
The biggest reason I've been able to move so fast lately has been because of these guys. Best QA agent I've found, with deep runtime errors, preventing me from shipping critical bugs. Highly recommend integrating this to any coding workflow. DM @sachelik , tell him I sent ya.
Our secret sauce for finding "hidden bugs": Most QA/review agents explore your code on their own but can miss out on connections that matter. We pre-process your codebase, infra, past incidents, and features into a context graph that guides the agent. So instead of guessing, it follows a real multi-hop trail and writes the probe that actually catches the bug.
3
2
8
714
AI will need you to get more technical (systems thinking), not less
If you’re not reading the code, whether explicitly or through agentic inquiry, one or more of these is true: ○ You’re a beginner ○ Software is throwaway ○ You’re prototyping ○ You have no users / revenue ○ You’re taking on debt & risk ○ Your problems are basic And btw. All of this is fine. But the reality is that models are still not at the “full autonomy” stage yet. They make rookie mistakes, they go down bad architectural paths. I just had the best model in the world add a nonsensical 700ms delay to “settle” something and it told me “you’re right, I was cargo-culting” 🤨 I am on the camp that this need will diminish more and more. Most code is indeed going to be assembly-like. But we also have the global internet and software infrastructure riding on these models and narrative, and we have to respect that.
4
533
welcome to usejina.com 🫡
Reviewing code will evolve just like writing (generating) code does / did. I don't see us review code the way we did in the past. And no, that does not mean that you just "yolo ship it". I don't know how new code reviewing processes will look like exactly. Maybe a combination of AI-generated summaries (markdown, HTML reports with nice, digestible structures) that include key code snippets / diffs with links to the relevant source code files. Maybe something else, I don't know. But I don't think the solution is to have a human skim over thousands of lines of changed or added code. Just as we're moving to a world where we trust research summaries without reading every source in full manually, I think we'll move to a similar model for coding. Just that there, we also have automated tests, the AI testing apps (if possible), us spending more time on defining specs and plans and workflows etc. So planning work upfront (designing a system, the code architectures, key decisions etc) + improved review processes thereafter.
1
5
510
Our secret sauce for finding "hidden bugs": Most QA/review agents explore your code on their own but can miss out on connections that matter. We pre-process your codebase, infra, past incidents, and features into a context graph that guides the agent. So instead of guessing, it follows a real multi-hop trail and writes the probe that actually catches the bug.
1
5
1,099
In SF for a week from July 25th to Aug 2nd Any recs for awesome AI folks to meet?
1
2
314
“Private AI” we all know what that means …
1
317
"Damn, I really love that Jina bot. It's even cooler than the Codex one" That was one of our users' first reaction. Jina behaves like an expert QA engineer. It iteratively explores and actually executes on the risky areas of every PR. A planner agent identifies areas to investigate, and investigation agents fan out and run through all of them in parallel. And after the first round, the planner goes through the results and decides where to dig deeper and what else to explore. That's the iterative execution loop that lets us keep finding deeper, higher impact issues.
2
2
7
730
Krish Chelikavada retweeted
An initial startup idea can't usually be both grand and precise. In practice they're usually either grand and vague or precise and small. Precise and small is better. You know who your initial users are, and you expand outward. With grand and vague you can't even get started.
226
428
72
5,604
417,014
Labs: Build as many usecases for our models Apps: Build the best experience for our usecase Implication: Labs - good for many things Apps - great at one thing
2
313
Krish Chelikavada retweeted
very good QA system because of its runtime eval. def recommend.
Jina is a fully autonomous AI QA Engineer - Maps your code, infra, and issue history - Traces the impact of any code change - Runs your app to catch and fix issues Jina outperforms static review agents, even catching what your engineers miss. Sign up for access below
1
1
3
979