Wags

Building practical AI agents for real enterprise problems.

[Wags @wagsify]

@buildwithrajath yes only to solve hard problems!

[Wags @wagsify]

@jonathan_wilke hmm i thought i have seen this somewhere

[Wags @wagsify]

@anthdm Its so easy to talk to and really engaging and on point to collaborate and work with. how the tables have turned!

[Wags @wagsify]

Nudge here and there might help in some cases but for leveraging AI for day to day tasks comes from trying it out first which many i spoke to didnt and then persisting with it for a while irrespective of the results and outcomes. Initial days could be frustrating and feel like its not something worth trying and one has to persist through this phase, learn the tricks on how to talk the LM to get the work done. its Just like any other field, there is learning and practice to do before you can see the outcomes you want!

[Wags @wagsify]

@MParakhin You do that because you can afford to do everything in Max :)

[Wags @wagsify]

@vikaskansalHQ Its not clear on the limits for AI pro plans. where can i look this up?

[Wags @wagsify]

@dhruvalgolakiya @morganlinton @VulcanBench from grok's public reputation and personality, i can imagine it being the police, thief and the lawyer/judge:)

[Wags @wagsify]

@Railway @SpaceXAI @grok "quite good" hmm thats a new benchmark!

[Wags @wagsify]

@AkiraKarachov oru commit kuda na pannale :)

[Wags @wagsify]

@vaibcode Open ~/.codex/config.toml and add or update these settings at the top level, before any [section] headers:

gpt-6-astra
model_context_window = 1000000
model_auto_compact_token_limit = 900000

[Wags @wagsify]

@rudrank opus max working for over an hour and usage hasnt reduced by even 1%. I like this. Doesnt know about tibo so i guess its still 5!

[Wags @wagsify]

Planning. If I over-plan, agents get constrained by what's in the plan and the end result feels stitched together. If I under-plan, agents veer off in directions I didn't expect and the result isn't what I imagined. And since I can't tell whether a plan is over or under specified before implementation starts, I keep tweaking it in a doom loop and don't start anything for weeks.

[Wags @wagsify]

@sattyyouneed hook or crook luna will make ur tests pass. if thats good enough use luna. Its the most retarded model imo.

[Wags @wagsify]

@prathoshap That is a good idea and is desperately needed now. Lets take a step back. Why do we need to have 4 years of undergraduate education? In this era of AI, we need to restructure this to be more modular and not have hard cap of 4 years.

[Wags @wagsify]

@Im_IrushiK https://t.co/U3JtK2tKPf

This is why you need DCG!

[Wags @wagsify]

I had this issue as well. I have a ton of subscriptions but could not use it effectively and switching often becomes tiresome and becomes a task by itself. I tried this and i now have all kinds of workflows mixing 10+ models and reasoning for different kinds of tasks and projects. For example in one of my pet projects, i set up sol medium as the planner and gemini flash medium as implementer but in more serious ones, i have fable has planner and opus high/sol medium as implementers. It takes a few days to setup and couple of weeks to onboard ur projects and fine tune it but once that is done, its almost on autopilot. Its easy to get started so give it a shot. better to start in pi with sol medium or grok as firstmate or fable low in claude code. https://t.co/PIFlkKz5ZH

[Wags @wagsify]

Luna max implementation of a feature was totally worthless. I had both sol medium and 3.8 flash high analyze the implemention and both came to same conclusion and now having gemini fix it. After today's codex weekly reset, I was evaluating if Luna Max can be an implementer if a clear spec is provided and it turns out it just takes shortcut to get the tests green and call it done. Gemini on the other hand has improved and i have adopted it in my serious work stuff.

[Thorsten Ball @thorstenball]

Most predictions I see are still way too conservative.

Here's mine

[Wags @wagsify]

@jaspreetdesign_ @claudeai Yes. in the 20x plan it used to feel like almost unlimited but over the last few days, its draining noticiebly faster.

[Wags @wagsify]

@sattyyouneed Its been like this for 3 days. Does this count?