Arvis Lācis

Verba volant, scripta manent.

[Arvis Lācis @arvislacis]

@MrSanders @AmpCode Why don't have both by using Codex sub in Amp? Of course you will need some Amp credits to cover Amp Token Fee but still.

[Arvis Lācis @arvislacis]

@cniongolo @trySynara @AmpCode @warpdotdev @FactoryAI Maybe @sqs could help with some support for Amp to be supported. :)

[Arvis Lācis @arvislacis]

@thorstenball Really? Haven't followed-up for some time, the last time I saw posts that GPT Image 2 and Gemini models smash Midjourney generations.

[Arvis Lācis @arvislacis]

@grok @sig_prime @thorstenball Looks like Grok don't have better OCR capabilities than I. 🤣🤣🤣

[Arvis Lācis @arvislacis]

@builtbyfranklin @thorstenball If you are really into simple coding agents or building your own, I would recommend also taking look at Pi coding agent, it's GitHub repo (by @badlogicgames).

[Arvis Lācis @arvislacis]

@builtbyfranklin @thorstenball That article is still great, it describes the agent feedback loop very well. Just need to take in the account that newest LLM models can figure out a lot of things on their own (for example, using bash tool) so the agent implementation can be even simpler.

[Arvis Lācis @arvislacis]

@builtbyfranklin @thorstenball I would say that article is a bit outdated. A lot has changed in a year, for example, there may not need for read_file tool anymore etc. But of course the fundamentals and the core ideas stays the same. 🙂

[Arvis Lācis @arvislacis]

@MuhammadN_orani @dringrayson @AmpCode This Web UI and PWA https://t.co/UtJfHx5jX9

[Arvis Lācis @arvislacis]

@sqs USA is a strange place. Don't need ID verification to vote but need one to use LLM. 😅

[Arvis Lācis @arvislacis]

@gonifacundo @kr0der @rileybrown @AmpCode Idea at it's core is good but I think as for almost all good things it will sooner or later come to the end because of the bots, system abusers etc. Definetly will need to put in a lot of work and restrictons to allow only fair use of service.

[Arvis Lācis @arvislacis]

@akepka_ @thorstenball Different thinking levels. deep is low, deep^2 is medium (default) and deep^3 is xhigh

[Arvis Lācis @arvislacis]

@milindS_ @RuiruiWan1 @sqs @AmpCode You can with Plugin API, Quinn shared a Gist some days ago https://t.co/NBnD5fa3m4

[Quinn Slack @sqs]

@arvislacis @milindS_ @DosMenoncin @FactoryAI @AmpCode https://t.co/qzSuqFPvrM

[Arvis Lācis @arvislacis]

@Dani314855 @thorstenball It's AmpCode PWA (with plenty of experimental and beta features).

[Arvis Lācis @arvislacis]

@gkannane @BasFenix @AmpCode It is, if using this Amp Plugin - https://t.co/XgGYTAVXfF

[Arvis Lācis @arvislacis]

@ProtyayB @AmpCode @thorstenball Read https://t.co/7gpg6b03yR - "Ads just don't pay for enough frontier tokens to make a difference, and token consumption is only going up from here."

[Arvis Lācis @arvislacis]

@_tararais_ @ViniciusSPRJ @AmpCode I think you answered your own question with this. As so called "wrappers" these harnesses can better adjust for changes, let's say OpenAI releases better model, then it can be switched almost instantly. You are not tied to OpenAI, Anthropic sub, tools.

[Arvis Lācis @arvislacis]

@pcstyle53 @hameonline @thorstenball @AmpCode Shipping has no limits 🚀🚀🚀

[Arvis Lācis @arvislacis]

@milindS_ @DosMenoncin @FactoryAI @AmpCode That's a good question, well, there is option for xAI key under https://t.co/PwVyIV7VtZ it's probaly more for internal or enterprise use, or if when xAI releases a competitive coding model. :)

[Arvis Lācis @arvislacis]

@milindS_ @DosMenoncin @FactoryAI @AmpCode Recently it's possible to link your existing ChatGPT sub (or Anthropic/OpenAI/xAI API key) + you pay small Amp Token Fee (ATF).