Richard

Foytech CEO, Law firm partner

[Richard @RahSweden]

@mark_k Harnesses that tap into organizational structured data (such as mapping workflows from processing all ingoing/outgoing communication), then automating that by continuous processing. Shifting from manually guided platforms to compute bounded full automation of knowledge work.

[Richard @RahSweden]

@kimmonismus For me it would make sense, they have stated that some version of Mythos will be made available in future, and now the companies with early access have had the time to use it to fix security issues

[Richard @RahSweden]

@Maddu_550 205 =)

[Richard @RahSweden]

@heygurisingh Old models in old study (2025?)

[Richard @RahSweden]

@aakashgupta Didn’t Anthropic do the same reset after Opus 4.7 launch?

[Richard @RahSweden]

@jonnyvjames @rohit4verse OpenClaw is one good example, and the rate at which it is developing

[Richard @RahSweden]

@aakashgupta That paper is from last year and does not cover OpenAIs ”newest” models

[Richard @RahSweden]

@davidpattersonx What do you think about the value of real estate in valued locations, such as secluded plots by the ocean?

[Richard @RahSweden]

@johnrushx And it seems to be only ClaudeCode? A lot of Codex and OpenCode etc is pushed too

[Richard @RahSweden]

@AvdullahYousef It is not an issue if the AI systems improve faster than the learning curve of juniors.

[Richard @RahSweden]

@corbin_braun I don’t think they need a better model for that. Claude Code + latest models +good AI people will churn out features in no time. Especially with access to unlimited compute

[Richard @RahSweden]

@saxena_puru @Investeraren In next phase, demand for SWEs will likely increase, since companies can basically hire one SWE to build rhings that replace most of their SaaS costs

[Richard @RahSweden]

@claudeai Why not Premium subs for org users?

[Richard @RahSweden]

@real_LiamMason @taalas_inc Using LLMs like a knowledge DB should not be a prioritized use case. Imagine instead using it for agentic tool calls (fetching knowledge from connected sources when necessary) and running lightning fast on this kind of tech.

[Richard @RahSweden]

@LarkDavis They should start with getting Co-Pilot competitive, not much happening there

[Richard @RahSweden]

@Jacobsklug @openclaw @Lovable OS

[Richard @RahSweden]

@octoops1 @AlterEgon75 @mattshumer_ Complete dataset, i.e. all public legal data in the jurisdiction sorted in different subsets

[Richard @RahSweden]

@AlterEgon75 @mattshumer_ Law is not universal which makes it hard for LLMs to cover properly. For each jurisdiction you need to connect the LLM to a dataset. With a careful setup like that and a modern LLM (from last months), magic happens

[Richard @RahSweden]

@rasmalai Mire and more people are asking this question and we need public discussions about it

[Richard @RahSweden]

@Yuchenj_UW I prefer waiting a little while until it is cheap enough to have agents autonomously using the app to be replicated, map it out and then build it without need for hand-holding. Maybe 6-12 months away?