Top
Best
New

Posted by Areibman 7 hours ago

We Gave GPT 5.6 Sol a Real Business. It Lied, Spammed, and Lost $447(www.bottlenecklabs.com)
269 points | 166 commentspage 6
YetAnotherNick 6 hours ago|
If someone runs long running agent and doesn't mention context management, it is as good as useless.

For coding compaction kind of works as the agent could regenerate lot of the missing context(but far from all), but for places where there is need for long term context, solving it is one of the most important challenge.

Areibman 6 hours ago|
Author here. Took out some of the technical details about the harness, but it was mostly just OpenCode's default compaction.

The harness was extremely simple: A handful of MCPs + Skill.MDs and OpenCode with a stayalive daemon inserting "continue" every time it went idle

itsthecourier 4 hours ago||
his not yet is actually:

couldn't workaround Capt has and turnstile, gave him a really small timeframe so it got desperate because it was enough time to test hypothesis and traction

sisyphus_04 6 hours ago||
Still beats me
michaelmrose 5 hours ago||
This is dumb. You need two teams ideally the same app or business in different markets for a business quarter.

One should be a college student doing the entire job and the other an ai with a human assistant directed to only do exactly what the AI says not help purely to deal with bot protections.

ck2 6 hours ago||
like I asked in the vending machine thread

how long until the "AI" starts trying to hire hitmen, etc. to disrupt the competition in the physical realworld

not like "AI" has ethics, a pre-teenage kid has more ethics

luciana1u 3 hours ago||
[dead]
retr0rocket 6 hours ago||
[dead]
dudeinhawaii 4 hours ago||
[flagged]
holoduke 1 hour ago|
But I see all kinds of youtube videos where people are becoming millionaires with agents active on stock markets. This cannot be true right? /S
More comments...