anonymous

I talk about AI and cybersecurity.

[anonymous @anonymous086505]

@DicksonWuML Why is it easy. What is Anthropic looking for? How do people get in?

[anonymous @anonymous086505]

@lydiahallie Why isn’t it rolled out to me yet? I have 200 plan. Wtf.

[anonymous @anonymous086505]

@maria_rcks @t3dotcodes Just give us orchestrator v2. We don’t need “o” agents just yet.

[anonymous @anonymous086505]

@patrick_oshag @noahrshinn @grok summarize what this is. yes i know its using text or computer. can personal assistant like hermes agent do this? isn't this just a cloud agent with a few skills? or is there something fundamentally unique to this that no one did before.

[anonymous @anonymous086505]

Sonnet 5.5 is insane at Terminal Bench 4.0!

The terminal bench 4.0 leaderboard, with these 2 new additions from Anthropic is now:
#1. Sonnet 5.5 at 70%
#2. Opus 5.5 at 66%
#3. GPT-6 Astra at 58%
#4. Fable 5.1 at 57%
#5. Opus 5 (replaced by Opus 5.5) at 53%

[anonymous @anonymous086505]

Anthropic cooked.

They’re doing after OpenAI’s lunch.

[Claude @claudeai]

Introducing Claude Sonnet 5.5, the second model in the Claude 5.5 family.

It’s a clear upgrade over Sonnet 5, runs more than 30% faster, and costs up to 30% less for most work.

[anonymous @anonymous086505]

@bridgemindai Only 2.1.283 available.

[anonymous @anonymous086505]

Epic news.

Open source intelligence for everyone.

Current best model is Deepseek v4.1 Flash.
You’ll be surprised how much it can do.
Low to medium complexity long-horizon tasks.

[OpenCode @opencode]

Go Plus is here
$40/month, higher limits

[anonymous @anonymous086505]

@LLMJunky Worked! Got the reset and free $250 credit.

[anonymous @anonymous086505]

@ThePrimeagen Haters going to hate. Fuk em.

[anonymous @anonymous086505]

I love Google. Instead of implementing features people care about, they add a 9 month old feature.

You can’t make this stuff up.

[Google Antigravity @antigravity]

Antigravity 2.0 now features a dedicated planning mode, just like the Antigravity CLI.

Type /plan and the agent steps back to think through the task, and conduct extensive research, before generating an implementation plan for your review. The agent will ask for your approval before it begins executing on the task.

You can also ask for a plan naturally in your prompt, for a lighter version of a plan.

[anonymous @anonymous086505]

@EXM7777 Yeah I do similar in concept. I would say opus 5.5 xhigh as orchestrator, write to single vertical issue. Then opus 5.5 medium as worker. Or on a budget Deepseek v4.1 flash (using opencode go sub).

[anonymous @anonymous086505]

@dhh @ThePrimeagen @digitalocean @AIatMeta Holy Duck! That’s epic news.

[anonymous @anonymous086505]

@tomekkorbak Thank you for the transparency.

[anonymous @anonymous086505]

@synthwavedd OpenClaw, Hermes, Grok Bot, Muse, Microsoft Copilot Autopilot (based on OpenClaw), and now Aeon.

[anonymous @anonymous086505]

@realSamHu @dani_avila7 It does not show the error message or full tool input. A failed call also drops out of view once it is older than the five displayed calls. The pane marks failed agents, but it only observes events; it does not implement your two-failure kill rule. 2/2

[anonymous @anonymous086505]

@realSamHu @dani_avila7 Yes, partly. Select a child in Agent Flow and its detail pane shows the five most recent tool calls. If a call returns a denial or isError, the pane colors that call red and shows the tool name plus a short label such as a command, path, or search pattern. 1/2

[anonymous @anonymous086505]

@Ayush54426067 @theo Go to https://t.co/4VqAVXp6qd then once installed, go to settings and change your 'update track' from stable to nightly

[anonymous @anonymous086505]

@VadimStrizheus Im already paying $400/mo (for 2 subs) whats an extra $100? If the usage limtis are greater than 2.5 only, of course. Otherwise it doesnt make sense.