prathap

swe @intainft . figuring things out

[prathap @Prathaptwt]

Trying to fill my timeline with people who share their best technical learnings on

- Distributed systems
- AI & LLMs
- Backend & system design

want to be around people who build cool stuff and share what they learn

[prathap @Prathaptwt]

@typesafeai 878 calls for 664 pages, 161s. 183ms a call.

168 docs came back above 0.90 confidence all 168 correct. the one mistake across 664 pages came back at 0.67 and got flagged.

that's the bar for unattended: not zero errors, errors that raise their hand.

[prathap @Prathaptwt]

benchmarked @typesafeai Jev on a production document classification problem, scanned loan packages (35–254 pages), classifying docs into 22 types.

The results were impressive:
• 99.8% page accuracy
• 100% boundaries found
• 11.2s / $0.023 per package
• 664 pages, 174 docs

[Diogo Almeida @CompleteSkeptic]

After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?

I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev

• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions

AFAICT the shortest path to AI-based economic revolution

[prathap @Prathaptwt]

@typesafeai then one Choice per finished doc, sent alone first time anything sees a whole document at once. fixes what the page pass got wrong, and gives the confidence score.

plus a Noul: "do these pages really belong to one document". without it nothing could contradict a bad boundary.

[prathap @Prathaptwt]

@jerryjliu0 excited to test how well it performs on document classification and confidence scores for extracted fields

[prathap @Prathaptwt]

@dotpem @typesafeai Hi, can I get access plss?

Relay excited to test it on document classification use cases and love explore it's potential in documents processing

[prathap @Prathaptwt]

@justKDeng excited to test it on asset-backed loan/lease document classification and explore its potential for document processing

[prathap @Prathaptwt]

Someone from @typesafeai could you give me access to jev?

Relay excited to test it on document classification use cases and love explore it's potential in documents processing

[prathap @Prathaptwt]

@ellen_in_sf Literature review

[Arpit Bhayani @arpit_bhayani]

With agentic slop, we are trading software reliability for shipping velocity and calling it progress. It isn't. Systems are more fragile than ever, and engineers building them no longer trust their code to hold up in real-world edge cases.

I am pro-AI, but this will backfire - big time.

[dax @thdxr]

mfs making a one line code change by writing a sentence into their agent and then posting about 10x productivity gains while it runs and bills them

[prathap @Prathaptwt]

@TEJAS5A65 @kirat_tw @100xDevs done. time to grind for the next 6 months 🔥

[prathap @Prathaptwt]

@hardikcoder @kirat_tw @100xDevs Done from my side. We’ll grow together

[Ajeet ( opensox.ai ) @jackedAJ]

now you can login https://t.co/77Pw1PRIlW using github 🚀

[prathap @Prathaptwt]

@ajeetunc Hehe, that’s me! ❤️