CTO @oruel_inc (EF S26) resident @lossfunk // ai research @iiit_hyderabad // recipient @aigrantsindia @3fdotVC @mercatus
- Tweets 1,590
- Following 35
- Followers 2,860
- Likes 6,830
chastronomic retweeted
Today we're launching @tryAlanAI so developers can stop babysitting their coding agents.
We kept running into the same problem ourselves.
Even as agents got better at writing code, we were becoming glorified button-pushers for our own AI: starting every run, feeding them context, unblocking failures, and checking their work.
We built Alan to change that.
239
103
506
139,712
my friend just interviewed for a bunch of startups for work. safe to say he's feeling extremely happy?
14
799
chastronomic retweeted
Congratulations to the 20th cohort! 🎉
Emergent Ventures India, Cohort 20. Congratulations to all the winners.
marginalrevolution.com/margi…
4
23
1,358
Emergent Venture Grant story time: we first heard about Emergent Ventures during my time at @lossfunk. since our startup idea (@oruel_inc) was at a very early stage, sourcing grants was extremely important.
i applied for EV 14 times in ~6 months.
the 15th time, I get a response from @srajagopalan and team that we got the grant. this grant helped us through some really tough times and kept our dream alive.
so, thank you @srajagopalan and team. this was amazing!
@prb5h @nihipap
Emergent Ventures India, Cohort 20. Congratulations to all the winners.
marginalrevolution.com/margi…
4
2
86
4,064
the first time you cook for yourself in a new house feels amazing. you chase that feeling everytime you cook there again
1
7
676
one week you're running the most ambitious startup in the world, and the next week you get called replaceable. all part of the game at this point
1
19
766
chastronomic retweeted
they should let me parachute out of the plane when it enters blr because why am i travelling 2 more hours after the flight
3
1
34
571
"i would rather remain a mere follicle in the body of society than become a monument to its folly"
6
471
i remember when I was 18, and a friend of mine introduced me to her work. immediately asked her if she has a nobel prize. this is amazing to see :D
7
487
Hope you caught @oruel_inc at AIM's Cypher 2026, Bangalore.
We defined 'Backend of the Frontier' and elaborated on how the space is up for grabs in several niche industries.
We are Oru'el. We build the next-generation of Infra & Data layer for Physical AI
cc:@chastronomic @prb5h
1
6
366
the usual Lossfunk stacked Ws. thanks for the @oruel_inc mention :)
Some exciting updates!
A bunch of @lossfunk residents have raised funding + a few Lossfunk incubated companies have received $25k each in Google Cloud compute credits. More details below 👇
---
@HariAyapps, who's building @archgen_, got into YC W26 🥳🥳
Residents who started building at Lossfunk got into @joinef S26: the @oruel_inc team (@chastronomic, @prb5h, @nihipap) and @suryamaddula_ building @WhisperwaveAI
Former Lossfunk researcher @harshablast started Dirac Robotics as part of @joinef S26
---
A total of $175,000 in compute credits has already been disbursed to 7 Lossfunk-incubated companies, and 9 more are in the process of receiving them.
Thanks, @googlecloud, for the credits.
• athenaagent: @sachdh
• attenor: @ViraajG1
• autr: @V_J_S_1
• fpvlabs: @levelheaded_94
• grit: @timbresociety
• prometei: @navbenny
• scoutfox
1
2
387
We’re HIRING Research Engineers & Research Interns at @oruel_inc.
We're building the infra layer for robotics and physical AI. Looking for strong research and engineering experience in multimodal learning, vision models, and robotics.
The work would be pretty straightforward: Read papers. Design experiments. Build and validate methods. Ship them.
The quality you bring matters more than anything else. If you're the kind to move fast and solve hard problems, you might be a fit.
Full-time + internships are open.
Apply here: oru-el.notion.site/3f2508f77…
8
14
215
14,677
chastronomic retweeted
catch Priyanshu (@prb5h) (Co-founder, CEO at @oruel_inc) at the @Analyticsindiam's Cypher 2026 event about the Backend of the Frontier.
@nihipap and I will be there at the event too, so if you're building at the intersection of Physical AI, Robotics and Data for Physical AI: DMs are open
3
14
654
this is actual Astra level code by the way. goes against the first lesson of cs fundamentals
2
6
668
fork found in kitchen
More pre-training doesn't necessarily mean better generalization?
This new paper discovers mode-hopping, where LLMs repeatedly switch between shallow pattern-matching and actual generalization, even while training loss remains stable.
For example, OLMo3-32B went from 81% accuracy to 0%, then back to 81.7% within just 40B training tokens.
On top of that, selecting an earlier 4.5T-token checkpoint instead of a 4.9T one improved GPQA transfer after math fine-tuning (36.3% vs 29.8%) and robustness to alignment attacks (53% vs 21%)
alphaxiv.org/abs/2609.33150
7
341
chatgpt needs to work on getting their stop response button to work properly before thinking of releasing their frontier model smh
2
7
507
