Two commands that I'm using a lot when coding these days:
/sitrep
- Mid-session situation report while work is still ongoing
/bluf
- Bottom line up front after real analysis �� reorient a cold context switch
As a coder I've seen AI code go from slop > useful > truly tasteful over the course of the last 18 months.
I think that's why so many of us are bullish on AI being able to produce tasteful output in other arenas in the near future, and why it might be hard for others to believe.
@ArchiveExplorer Jeff handles this failure mode explicitly: every blocking finding goes to a fresh critic and can be dismissed only with concrete code evidence. Uncertainty keeps the blocker alive. The builder never signs off its own work. https://t.co/9d7eStyA67
@Tech_girl I find that with the right system in place the outputs can be very reliable.
- Questions up front
- Tests before code
- Agents don't review their own code
My system is called https://t.co/awf5GDaF0s, but there are other systems out there that are good.
@unclebobmartin Grok 4.6 works very, very well with a solid system like https://t.co/rMWIO32OR9 around it to enforce good behavior, like agents not checking their own work.
I default to Grok 4.6 these days.
It's quite possible that the $300 SuperGrok Heavy plan will replace both my $200 Claude and OpenA plans.
I was getting good milage from Grok alone without realizing that the Cursor Ultra plan was included!
Having used Grok @bot for about 6 hours I can say that we're definitely one step closer to Spike Jonze's "Her" (2013). Hopefully just the good parts.
It's a good watch in 2026: https://t.co/yyS98P6zgj
@TechByTaraa I see a lot of Rust answers, but Rust is really a popular language these days.
Clojure is more of a hidden gem. I don't use it day-to-day, but it changed the way I think about software. Like learning Greek or Latin.
@bossriceshark I'm getting insanely good value out of it. I'd say it's better than Opus 5 but not Fable 5. Grok 4.6 XHigh is on par with GPT 5.6 Sol High in my experience.