Jason Broander

[insert self promotion...] Most importantly: husband, father, citizen, patriot. Romans 8:18

[Jason Broander @jbroander]

https://t.co/057mDTrRj1

[Jason Broander @jbroander]

@danielcronauer @CoFoundersNik Me too. My income is high, but our household is not top 1%.

[Jason Broander @jbroander]

@mtkonczal Do the same chart wirh four plus.

[Jason Broander @jbroander]

@Afinetheorem The market for AI may be capped by human talent you’re saying? Because it’s now the scarcer resource? If so if you don’t hit AGI it means we have over allocated capital to AI. It also means high IQ gets ground into a pulp trying to run faster until equilibrium is reached

[Jason Broander @jbroander]

@nicoypei These are symptoms of a bubble.

[Jason Broander @jbroander]

@confucius_dr @RickolaTesla @usr_bin_roygbiv Yeah get everyone a 2k gpu consumer laptop with qwen code on it and they’d be as good as last fall, which is better than this situation by far

[Jason Broander @jbroander]

@pitchprofiler Well said

[Jason Broander @jbroander]

@PermitZIPhvac They’ll fire a couple of peons who were following a playbook handed to them no doubt, and change nothing.

[Jason Broander @jbroander]

@jon_stokes Judgement should be added, part of a breakdown of execution quality (not just speed). The models prove if you can’t ask the right questions you don’t get the right answers. That’s another immutable point. What questions should you ask can’t be solved by the model.

[Jason Broander @jbroander]

@ericlu Count, group by region

[Jason Broander @jbroander]

You might be mistaking “top reps” with operating in an environment where they are mostly order takers. You do need to ask for the sale one way or another otherwise objections don’t come out at the right time. However you should use the right phrasing to march the audience and keep tension dialed not too high. The direct phrase you mention might be too much tension, but not always. As some had said here, some hiring managers expect this approach, so you need to factor the market into your evaluation criteria.

[Jason Broander @jbroander]

@ByrneHobart This would be an interesting discussion to continue. I think a good way to frame would be thinking about which of the three acts of the mind would we be asking the agents to help with (understanding (apprehension specifically)> judgement > reason)

[Jason Broander @jbroander]

Now we’re getting to the really important questions

[Byrne Hobart @ByrneHobart]

Have any of the labs tried to do in history what they've done in math, like digitizing the relevant documents, having lots of historians in the loop, and throwing a few billion tokens at a question like "why did the industrial revolution happen when and where it did?"

[Jason Broander @jbroander]

@RylanSchaeffer Yeah that’s interesting. Say more?

[Jason Broander @jbroander]

@p_millerd @Jack_Raines @bot No kids yet?

[Jason Broander @jbroander]

@tszzl @shakoistsLog @johngfriedman Enhance

[Jason Broander @jbroander]

@theo What effort level are you using?

[Jason Broander @jbroander]

@conjibz @omarvvvr And get a free tier oracle cloud VM so you can run multiple agents from anywhere, controlled with /rc on your phone on the go.

[Jason Broander @jbroander]

@conjibz @omarvvvr

[ⓙⓘⓑⓞⓢⓢ @conjibz]

@omarvvvr tmux

[Jason Broander @jbroander]

I know a few weeks ago anthopic (might have been Boris himself) was recommending, and the data was showing, that running Opus 5.5 on medium got basically similar results to the highest effort level. I wonder what the details of the test framework, because I'm having much better results with the highest when I'm collaborating on a project. Running on medium just seems to regress into you micromanaging the agent step by step, vs highest it's sending out agents to and fro as it should.