Building AI-powered products 🚀 Currently developing Zeltrika, Roguelo & @BeeAgentsDev 🐝 Sharing the process: progress, demos & lessons.
Joined May 2026
- Tweets387
- Following171
- Followers117
- Likes469
Pinned Tweet
Looking to connect with people interested in:
• AI agents & automation
• SaaS & app development
• Building in public
Here’s my project: BeeAgents 🐝
Build specialized AI teams, connect them into workflows, and follow their progress in one workspace.
Bring your own models. Work locally or through the cloud.
More soon.
Honest question for people using coding agents every day:
How often do you read the full diff before you merge?
“Does it compile?” is only the first test. Does anyone actually enjoy using what the agent built?
Simon’s example is games, but the same question applies to product software: working code still needs to become something people want to use.
teapawt.org/simonw/status/21154447…
I think October represents another inflection point in capabilities for coding agents and the models that drive them: Claude Opus 5.5 can now build video games that are actually fun simonwillison.net/2026/Oct/2…
Banger PR
Summary, evidence, merge danger on point
A great example of my /pr skill in action
github.com/mattpocock/course…
Hello to everyone who followed along this week! 👋
I’m Adrián, building BeeAgents: a workspace for AI agents. I’m sharing what I learn along the way, including the things that break. 🐝
What are you building?
What happens to work when AI can do the thinking and robots can do the physical tasks? 🤖
Musk’s vision is an age of abundance: far more goods and services, broader access to healthcare, and eventually a “universal high income.”
That’s a prediction about the future. How those gains would reach everyone is the question that matters.
Would cheaper essentials change your life more than a bigger salary?
teapawt.org/Jeremy_Lefcn7l/status/…
🚨 Elon Musk has dropped an unprecedented bombshell—right inside the White House, in front of Donald Trump: 💥🏛️🇺🇸
“From now on, you will no longer need to work for any company... I will make sure everyone on Earth enjoys a celebrity-level lifestyle and income!”
Musk is completely challenging the traditional concept of government-provided “Universal Basic Income” (UBI) designed merely for survival, presenting a radical vision of a new form of capitalism.
According to Musk, how will everyone become wealthy? 👇
As AI and robots produce goods and services 24/7 without limits, the overall standard of living for all humanity will skyrocket—as if everyone were earning an astronomical salary!
Musk confidently declares: “Everyone will have access to better healthcare than I personally receive!” Through the combination of superintelligent AI and precision surgical robots, top-tier medical care could become available to everyone at virtually no cost.
AI will handle the thinking, while robots build roads and structures, grow crops, and manufacture products. With forced labor becoming a thing of the past, the phrase “I don’t have enough money to buy what I need” could disappear from the human vocabulary entirely!
While 99% of humanity worries that “AI will take their jobs,” Musk envisions a future where “an army of autonomous robots will turn every single person into a king—living a life of total ease!”
Bookmark this post to follow the transformation that may lie ahead. 📌
OpenAI just announced up to $500/month in API credits for eligible ChatGPT Pro and Business subscribers.
My backlog of “I’ll build this when I have API credits” ideas just lost an excuse 😅
teapawt.org/OpenAIDevs/status/2115…
100 followers. A small number, but it already means a lot to me. 🙏
When I’ve shared what I’m building, you’ve asked questions, suggested ideas and helped me see things differently.
I’d never really used X before. Finding fellow AI nerds who are happy to help without expecting anything back has been the best part so far.
Thank you. Glad I started sharing here 🐝
Adrian Lopez Segura retweeted
Building a project?
Tell us what it does and what you need help with.
Let’s connect and help each other out — try each other’s products, share ideas, or work through a problem together.
What are you building? 👇
Building a project?
Tell us what it does and what you need help with.
Let’s connect and help each other out — try each other’s products, share ideas, or work through a problem together.
What are you building? 👇
I’ll start 🐝
I’m building @BeeAgentsDev: a workspace for teams of AI agents, each with its own role, model and tools.
Still in development. The goal is to make coordinating agents easier, without losing track of what they’re doing.
What would you want an AI team to help you with?
An AI agent can pass every check and still get the work wrong.
A valid-looking result needs review, whether that comes from a person or another agent.
Long-running agents are only useful if they stay on task.
Google is positioning Gemini 4 Carbon around cybersecurity and long-horizon work. I'm curious how it handles an existing codebase when the plan breaks halfway through.
Gemini 4 Carbon is #1 on Artificial Analysis' Intelligence Index v5, but #2 on its Coding Agent Index.
That's the useful distinction: the best model depends on the job. Which one would you test first for building products?
Google’s Gemini 4 Carbon is the new #1 on the Artificial Analysis Intelligence Index v5, and second only to Claude Opus 5.5 on the Artificial Analysis Coding Agent Index
Gemini 4 Carbon scores 59 on Intelligence Index v5, driven by leading performance in complex reasoning, knowledge, and scientific tasks. It also has large gains in coding and knowledge work, particularly when compared to other Gemini models, but it is not leading in all these areas.
As with launch of Gemini 4 Argon, Gemini 4 Carbon is being rolled out to selected users and is not yet publicly available. Google is offering a 50% launch discount, with no confirmed end date. All our charts and figures use standard pricing for both Gemini 4 Carbon and Argon.
Key results for Gemini 4 Carbon with high reasoning:
➤ Gemini 4 Carbon leads the Intelligence Index v5, with the #1 score in 4 of the 9 Index evaluations: HLE-Diamond, Terminal-Bench-Science, AA-Omniscience, AutomationBench-AA. Carbon makes substantial gains elsewhere but is not leading in coding and knowledge work
➤ Behind only Claude Opus 5.5 on coding agents: With Antigravity CLI, Gemini 4 Carbon (high) scores 59 only 3 points behind Claude Code with Claude Opus 5.5 (max, 62) on the Artificial Analysis Coding Agent Index. Carbon ranks second on both Terminal-Bench 4.0 (64%) and FrontierCode v1.1 (51%), behind Claude Opus 5.5 (max) at 56%. Google has historically underperformed in this area and so this improvement stands out
➤ Improved agentic knowledge work, but not yet leading: On AA-Briefcase, Gemini 4 Carbon (high) scores 1609 Elo, a 126-point improvement over Gemini 4 Argon (high, 1483), but still behind models including Grok 4.7 (xhigh, 1638), Muse Wonder (max, 1659) and Kimi K3.1 Pro (1678). The largest gain comes from Presentation, up 293 Elo to 1598, alongside a 120-point improvement in Analytical Quality to 1671
➤ Lower Cost per Task than Claude Opus 5.5 for comparable overall intelligence: Gemini 4 Carbon costs $7.76 per Intelligence Index task, ~23% less than Claude Opus 5.5 (max, $10.04) but ~1.2x GPT-6 Astra (max, $6.25). It uses 123k output tokens per task, 29% fewer than Claude Opus 5.5 (max, 173k). At the same $4/$20 standard pricing, its Cost per Task is in line with Gemini 4 Argon ($7.68), while scoring 9 points higher on the Intelligence Index
➤ Leading AA-Omniscience v1.1 score, with low hallucination: Gemini 4 Carbon scores 54 on our recently updated AA-Omniscience v1.1, 6 points ahead of Claude Opus 5.5 (max, 48) and 8 points ahead of Gemini 4 Argon (46). Accuracy (% correct) rises 10 points over Argon to 58%, while its hallucination rate rises from 4% to 11%. This is still the second lowest of any model, behind only Argon, compared with 50% for both Claude Opus 5.5 (max) and GPT-6 Astra (max)
Key model details:
➤ Context Window: 1M tokens
➤ Multimodality: Text, image, video, and speech input, with text output
➤ Pricing: $4/$20 per 1M input/output tokens, the same as Gemini 4 Argon’s standard pricing, with a 50% discount. Cached input tokens receive a 95% discount ($0.20 per 1M)
Congratulations to the @GoogleDeepMind team!
👀👀
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!
Adrian Lopez Segura retweeted
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!
A Wednesday question for people building with agents:
Every bee in BeeAgents gets a role, a model and tool permissions before it runs.
Which of the three do you set first?
And which one do you end up changing most after the first run?
Robots in 2015 vs. robots in 2026 🤖
The progress in just over a decade is wild. What do you think the next 10 years will bring?
Adrian Lopez Segura retweeted
Prompt of the day:
Split this "one-way door" PR into two:
1. A PR that implements everything, except it can be easily reverted (a two-way door)
2. A followup PR (stacked on top) that actually goes through the one-way door
AKA expand/contract. Use feature flags as needed.
---
Lets you de-risk a huge PR by testing large chunks of it in production first.
Tibo just shared poteto.com and now there’s a whole playground of little characters 🥔👀
What do you think this is teasing? My guess: maybe something to do with AI agents or bots—but that’s just speculation.
