sreekanth850

4 ms·

1.
by sreekanth850 14d ago
I don't really see the breakthrough in Jev. Classification, scoring, routing and returning probabilities over predefined choices are all established problems. We implemented category routing in our own retrieval system in a slightly di
2.
by sreekanth850 16d ago
I wish this can be a extension of existing languages.
3.
by sreekanth850 18d ago
I'm doing this. After getting started with LLM coding, I became super interested in learning to code, just out of passion. I walked out of engineering thinking physics was elite, but now I understand how passionate I am about building
4.
by sreekanth850 18d ago
This is best thing to use for decision making, evaluation, classification. If I'm not wrong.
5.
by sreekanth850 21d ago
i don't need all this shit, can i have a modern browser that run in 10 mb ram and doesnt multiply the memory with tabs. something like kmelon but works on all sites.
6.
by sreekanth850 22d ago
I’m using dpupek/roslyn-code-navigator (codenav-mcp) with Codex.
7.
by sreekanth850 22d ago
Yes. and i find model makes less errors and reasoning the codebase well, especially when you do a large refactor.
8.
by sreekanth850 23d ago
i don't know if such hacks works, but in C# if you use roslyn mcp, you save a lot.
9.
by sreekanth850 23d ago
Pretty happy with Luna. We use C# and add roslyn compiler MCP and Graft MCP, its super efficient and like infinite usage on plus plans. Maintaining 3 Rpeo with size of 360 K Loc. And a dozens of smaller repo collction together exceeds 500 K
10.
by sreekanth850 25d ago
The problem is when people ignore or dismiss developments in AI purely out of insecurity or a sense of human superiority. Machines don’t have to be superior to humans to dramatically amplify what humans can do.
11.
by sreekanth850 25d ago
Many will disagree and say that anything done by humans is by far superior to anything done by machines. You don't want to correct them, do you?
12.
by sreekanth850 25d ago
A fuzzer is a tool. An LLM can decide when to use the fuzzer, interpret the result, switch tools, change strategy and continue toward a high level objective.
13.
by sreekanth850 25d ago
D o you think a machine gun is marter than humans? or a car is smarter than Human brain? Human doesnt need to test, if the outcome can be tested deterministically by harness. The model tries. The harness checks whether the expected outcome
14.
by sreekanth850 25d ago
I think people here still evaluating the model in isolation. It is the combination that matters, model + strong harness + tools + long running autonomy + memory + retries + parallel agents + code execution + credentials + access to real sys
15.
by sreekanth850 25d ago
Model doesnt need to. Human bran never do either. its the mix of Model + harness + tools that will become dangerous combo. See how coding chanegs when agentic harness released?
16.
by sreekanth850 29d ago
I'm not a neuroscientist, Our brains also rely heavily on prediction, using prior context and experience to anticipate what comes next, Also, people those who talk about the deterministic nature, LLM doesn't need to be determinis
17.
by sreekanth850 1mo ago
In my mother tongue, Astra means arrow that used to destroy enemy.
18.
by sreekanth850 1mo ago
Dear Google, Kindly make you chat window on the right side of vscode in antigravity extension, There is a reason others kept it like that. I can see the code and inspect the files changed while Agents keep working. its critical for me perso
19.
by sreekanth850 1mo ago
non one forces anyone to use something. just ignore and move on. The author had clearly stated that its 100% written by Claude and codex with proper prompting and spec document.
20.
by sreekanth850 1mo ago
Platform risk is real. long back we decided to build a review management platform with Google as main. After building and getting early customers, we changed our office location that triggered to re apply for google business reviews account
21.
by sreekanth850 1mo ago
He forgot to add Make no mistakes in his prompt.
22.
by sreekanth850 1mo ago
2 years before everyone was doing custom training. What happened to all those today when frontier models itself become more powerful than custom trained ones?
23.
by sreekanth850 1mo ago
It all started when OpenAI began focusing on Codex and coding while sunsetting Sora, which helped free up a lot of compute and resources. 5.6 was the final nail in the coffin.
24.
by sreekanth850 2mo ago
Yes, especially something that touches DB. Router and shell are stupidly simple and you get auto failover in some 30 minutes. That is something make me stick to mysql.
25.
by sreekanth850 2mo ago
How do you implement HA in postgres, i found MySQL HA stack pretty straight forward with Innodb cluster, MySQL router and Shell.
26.
by sreekanth850 2mo ago
They are Ex AMD employees. Nvidia tried back, but they rejected.
27.
by sreekanth850 2mo ago
visual design and polish is highly dependent on the taste and standards of each team. and we’ve never gotten output from any of the frontier models that we felt was good enough there. where business logic have clear pattern and algo to move
28.
by sreekanth850 2mo ago
We have the UI mainly for Admin console, demo etc. Its a b2b enterprise product, so we don't use any custom CSS libraries.
29.
by sreekanth850 2mo ago
Ai is weak for front ends. We do hand coding for all UI related things.
30.
by sreekanth850 2mo ago
AMD along with cerebras may probably compete with NVIDIA monopoly in near future. Also, NVIDIA will have competition form multiple companies. Just my prediction.

More ›