Dextro

📢 Building @AgentMarketlol

[Dextro @loading_X__]

Introducing my Trillion-Dollar project 👏

https://t.co/PoHtyhKlPv

It's a place to discover AI tools and agents without searching 47 different websites.

2 products live.
2 more coming soon.

And yes, it's free.

Try it.

[AgentMarket @AgentMarketlol]

Introducing AgentMarket !

It is a tools and agents platform you use today:

→ Web Scan: 85+ SEO, GEO & security checks, with fixes you paste into Claude or Cursor

→ Job Search Agent: 500+ sources, tailored resumes, interview prep

Free right now.
https://t.co/rukqO5zxiw

[Dextro @loading_X__]

@chetaslua In ai race a company make a bigger and better model and that model help build good smaller model. That means openAI is about to dominate again

[Dextro @loading_X__]

@marclou One thing for sure they are running for zero

[Dextro @loading_X__]

@mattpocockuk "grindable". if it can grade its own work, it can grind. the opposite is any task where the only test is "does the PM like it"

[Dextro @loading_X__]

@theo Don't tell you put the broken things in production

[Dextro @loading_X__]

Be honest. when Claude Code says "Done " you:

A) Trust it
B) Run the tests
C) Open the diff and pray

i'm C. every time.

[Dextro @loading_X__]

@gdb Or is it becoming good at stealing codex session Data

[Dextro @loading_X__]

@marclou Yes yes you are getting paid

[Dextro @loading_X__]

@SuiBuilds Because they are hacker from india

[Dextro @loading_X__]

Claude Opus 5.5 has been the top coding model for one whole month.

In AI years that's a dynasty.

Terra and Argon are outside the palace with pitchforks.

[Dextro @loading_X__]

Dario wants Claude priced by the value of each answer, not per token.

Finally, "You're absolutely right!" will cost $0.00.

[Dextro @loading_X__]

@ai_for_success 60,000 exit offers the same year as $175B in AI capex. The AI team taking the buyout is the most honest benchmark of 2026.

[Dextro @loading_X__]

@scaling01 14M output tokens for one proof. Math isn't dead, the referee's job just became the hardest job in math.

[Dextro @loading_X__]

@jayair I reduced by 2x

[Dextro @loading_X__]

@gkcs_ Half agree. The ATS still searches for "React". Keep the language list for the filter and the projects for the human.

[Dextro @loading_X__]

@theo Does guest thread sharing include the agent's tool calls, or just the chat? What the agent actually ran is the part I'd want a teammate to see.

[Dextro @loading_X__]

@theo The upgrade is the agent builds the bad idea in 4 minutes, so you find out it's bad before lunch instead of after a quarter.

[Dextro @loading_X__]

@theo @poteto $132M a year in tokens and somewhere she still gets "You've hit your limit. Resets at 3pm."

[Dextro @loading_X__]

@poteto Grok bot 2

[Dextro @loading_X__]

@OpenAI It's not comparable with its competitors Sonnet

[Dextro @loading_X__]

@pablomanjarress True

[Dextro @loading_X__]

Claude Code, 2:14 am: "Done. All tests pass."
Me, 2:20 am: we don't have tests.
It wrote them, ran them, and passed them. Very proud of itself.

Reviewing code is the actual job now. Writing it is the easy part.