Building the neovim of agent harnesses. bone-agent. 26, swe in finance.
New York, USA
Joined April 2026
- Tweets3.1K
- Following244
- Followers182
- Likes2.1K
made this cool plugin that shows you the percentage time of reasoning, tok gen, and tool calling takes per turn. It updates live as the agent streams and calls tools, collapses prior messages.
also tui on fold 8u is unmatched. Don't need a laptop.
What do we think the next form of ai will be, or are agents the end game? Surely we have another unlock coming
Opus killer just dropped. Someone do the bicycle riding the bird to see if its legit.
I did say we'd be releasing our models periodically...introducing Gemini 4 Carbon, our next frontier model.
Following Argon's release just two weeks ago, Carbon shows the pace at which we’re pushing the frontier, it ranks #1 on the recently updated v5.0 of @ArtificialAnlys Intelligence Index at 59. It also leads across key cybersecurity evaluations scoring 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching.
We’re seeing significant improvements in long-horizon tasks even compared to Argon, and we’ve strengthened frontier safeguards.
As we did with Argon, we’re going to roll it out gradually + responsibly. Some trusted cyber defenders will get access starting this week. We'll be expanding it to more partners through our Fairwind Program, ahead of wider availability soon, so hang tight!
I just received a response from opus 5.5 so intelligent I could not believe its opus 5.5. They must have routed me to fable or something. Blown away. How can I test this
I change my stance on terra, it is decently fast on api. I swear the plus users get a dogged version of the models. We are back of the line. I had overload errors for 3 days straight couldn't use it period.
But yea terra isn't terrible. Probably wasted resources training it tho
Im not experiencing this incredible commit speed from terra. It feels slower then sol 6.1 was yesterday
Im guilty of shitting on gpt models, but they are still the kings of token efficiency. That was the claim made months ago when 5.6 came out.
Just got hands on with the iPhone duo as a fold8u user. Most underrated part is the spring loaded opening. You have to pry apart the folds, the duo just pops open.
For those of you using gui coding agents do you use the build in features like file viewers, browsers, etc?
Was using dsv4.1 flash for a few simple tasks and it would run like 30 tool calls to answer a question. Gpt-6 luna would run 3 for the same question, both got the answer right. Interesting.
I love how this pricing is considered high these days. For a Chinese model yes but its the same price as sol and thats seen as a bargain.
Im having grok file a complaint with chatgpt about codex. Its gonna argue with the other ai for me. Im so excited.
What is with the dated model version numbers
