chalk on a whiteboard. a long session of talking through some really crucial things, that are being forgotten.
great watch youtube.com/watch?v=UsvSfP3I…
there’s going to be a generation of software kids who will grow up without knowing what a pointer is, how async-await works and what the hell is OOP even?
Cyclomatic complexity is a silent killer. bites you bad and you don’t even know.
Keep an eye out for it and untangle your decisions as much as you can.
5
14
422
Bas beizzati kara lo
This Diwali, a little more light from India's AI research community onto the global AI landscape.
Introducing the latest benchmark results from Alpie Seed, our 9B multimodal reasoning model developed at @169Pi_ai Research.
We believe the next generation of AI will be shaped not only by models that scale bigger, but by the research and engineering that make smaller models increasingly capable.
We put Alpie Seed through a range of demanding evaluations across reasoning, scientific knowledge, mathematics, coding and software engineering.
Here are some of the results:
→ 82.54% on MMLU-Pro
→ 84.32% on GPQA Diamond
→ 72.99% on LiveCodeBench v6
→ 61.20% on SWE-Bench Verified
The comparison highlights some interesting results.
Alpie Seed records the highest GPQA Diamond score among the models shown. On MMLU-Pro, it performs competitively with larger models, while its LiveCodeBench result is within striking distance of substantially larger systems.
Its SWE-Bench Verified result of 61.20% is another important milestone, reflecting its ability to work through real-world software engineering tasks. This evaluation used mini-SWE-agent without retries.
Beyond benchmark accuracy, we also examined inference efficiency.
In a separate profiling evaluation, Alpie Seed recorded:
→ 36.66 ms time to first token
→ 41.37 tokens/s generation speed
→ 9,796.22 tokens/s prompt processing speed
These measurements come from a separate profiling setup and should not be confused with the comparative benchmark results.
What makes this work meaningful to us is the process behind it: developing a compact model through supervised fine-tuning and reinforcement learning, evaluating it across diverse tasks, and making the resulting weights accessible to the wider community.
The comparison includes published results from other model developers. These evaluations were not rerun under a common harness, so differences in methodology and configuration should be considered when interpreting the numbers.
Built in India. Developed for a global research and developer community.
This is Alpie Seed. And this is one step in a much larger journey for 169Pi Research.
Explore the model:
Hugging Face: huggingface.co/169Pi/Alpie_s…
GGUF: huggingface.co/169Pi/Alpie_S…
Ollama: ollama.com/169pi/alpie-seed
Kaggle: kaggle.com/models/alpie/alpi…
Happy Diwali from everyone at 169Pi.
11
470
Last diwali, exactly on the day, was my final round of interview with my current company.
it’s been nothing but great times since then. always grateful for all the love from everyone here. Happy Diwali!! Best wishes to you all. 🪔
update: been over a month with this setup and it is seriously great. regular meaningful updates are only elevating this experience.
the fact that it has an inbuilt editor, diff viewer, browser, terminal/custom chat UI (also a phone emulator if you need it) and all of this is accessible from the iphone app too.
been really helpful for productivity
While I have loved my terminal based workflow with kitty + herdr, I recently moved to using orca (@orca_build)
and honestly, while I miss the level of control and customization I could achieve, this actually makes things really great to work with. managing projects and workspaces feels simpler, a built-in editor, browser, some neat skills alongside some neat integrations with github and linear.
allows setting up automations, easier checkouts, tracking tasks. all I'm looking for is a higher level of workspaces (project folders?) that allow me to separate out work and personal projects and some more customizations. But apart from these, super good stuff.
Give it a shot!!
2
11
275
been saying this a lot lately: write your evals from day one!!!
youll thank yourself only a few weeks later!!
17
417
how does one wake up and decide to do this. it’s insane to see it working, but to have even thought of this is crazy
we’re clearly not thinking far outside the box enough
Today at work, we all had a moment of excitement as well as anxiety. Some problems that required full scale ml expertise only a year ago are now being solved by agents.
there’s obviously the challenge of costs and scale, but intelligence is getting cheaper by the day. times are truly changing. It’s troubling because my value is shifting and I dont know where it goes next.
but at the same time, there’s also excitement in the air. it doesn’t feel like an end, it feels like a new beginning.
(sounds a bit too dramatic, but this was my genuine reaction today lol)
8
35
960
Aarno retweeted
Introducing Orcus.
No LLM API key required to make AI model requests.
Orcus uses your web credentials to access AI models available through supported provider websites and deliver responses to your applications through a familiar, API-style interface.
From model invocation and streaming to structured outputs, tool calling, and batch processing, Orcus brings essential AI integration capabilities into one developer-friendly SDK.
Explore Orcus: orcus.site/docs/gemini
















