@MrAhmadAwais@entrptaher@CommandCodeAI@opencode@MrAhmadAwais isn’t this misleading again. Remember you wrote a whole paragraph on Reddit claiming “our rates are ours, not a deepseek invoice” and we have our own infra, we’re not pass through of deepseek.? What happened to that.?
https://t.co/d7R3icqKy2
@JackWoth98 It's working great for me! however, it's making some weird assumptions/decisions on its own. Like, I added a feature, push the code to GitHub, and ask it to deploy - no issues. But then on the next run, it just decides to auto-commit and deploy without me asking.
@1kartikkabadi1 It’s definitely the Harness, not the model. had same issues with @CommandCodeAI. Try OpenCode; completely different experience. No hallucinations, just a plan and silent execution. Harness needs serious improvement on this front.
@SivaramPg@CommandCodeAI Understood, for me it's taking more turns than needed for simple tasks and I'm seeing hallucinations. I also recall being able to click on individual requests for cost and cache breakdowns, but that seems to be gone now.
@MrAhmadAwais It’s burning tokens like crazy, even simple tasks burns 19M tokens, I haven’t been able to produce anything meaningful and burned almost 70M+ tokens. You need to work on your harness.
@jayair Hey @jayair , I’ve emailed @opencode/@anomalyco for three days straight and still haven’t received any response. At this point, I genuinely have to ask: does your support actually exist? Because from the user side, it feels like nobody is responding at all.