Ram Navan
@ram_navan
Followers
246
Following
2K
Media
21
Statuses
991
Operating a small portfolio of software products.
San Francisco, CA
Joined June 2008
Maintaining context with multiple agents and multi-tasking is the new developer superpower.
0
0
2
Agent or Workflow - at least for now, is in the eye of the beholder
0
0
0
Tried same prompt with Codex and Claude Code. It was for building a new feature in existing large codebase that called for understanding nuances of how various components and services work with each other. Claude made mistakes. Codex one-shotted a working solution. Impressed!
1
0
1
hey @sherwinwu 👋 - Could you look into the performance issue with responses API? Context here https://t.co/7RLCnHl49x 🙏🙏
community.openai.com
I posted this as indirectly related in another thread, but the issue seems specific enough to warrant its own thread. On Chat Completions, GPT-5 (minimal reasoning) is averaging about 5-7 sec for me,...
0
0
1
Responses API from OpenAI has 2-4x higher latency than chat completions. Disappointing. Hope they fix it soon
0
0
0
If you are switching from 04-mini, GPT-5-mini is a good option. GPT-5-nano is bad even in some simplest use cases and the experience feels very different when migrating from 4 family of models.
0
0
0
Chatgpt is much better in emails than Gemini. Gemini adds a lot of fluff
0
0
0
One thing almost all LLMs are bad in is npm package recommendations. It usually has high confidence in recommending packages that are abandoned.
0
0
0
Claude code has improved a lot in the past few days. Especially in its planning before implementation. Good job @AnthropicAI !
0
0
0
not good claude "The crash was caused by a reference to a non-existent variable xxxx in the comment."
0
0
0
The fact that Python is the standard for machine learning is a serious indictment of the field’s engineering standards
794
689
15K
Comet tip: Shift + Enter in address bar to get google search. Sometimes you just want to get to a website and google search results UI still bets for this use case
0
0
0
Claude Sonnet 4 now supports 1 million tokens of context on the Anthropic API—a 5x increase. Process over 75,000 lines of code or hundreds of documents in a single request.
725
1K
15K