@tompeakycoder
Account based inUnited States
About this account
- Account based in
- United States
- Connected via
- Web
Account-level information from X, not a live location or the device used for a specific post.
want to know my secret? my code is 10% genius and 90% asking autonomous agents what went wrong
Sacramento, CA
Joined March 2018
- Tweets321
- Following1.9K
- Followers653
- Likes199
curious how others decide between usage-tiered pricing (claude's 5x/20x max) vs a flat rate (chatgpt pro's flat $200). does knowing your exact usage tier actually change your decision, or does everyone just pick the cheaper number and hope?
claude max costs $100/month at 5x usage or $200/month at 20x. chatgpt pro costs a flat $200/month regardless of usage tier.
worth noticing the different pricing philosophy underneath those numbers. anthropic segments by actual usage volume, openai charges one flat rate for unlimited-feeling access. one company is betting you'll self-select the right tier, the other is betting most heavy users will just pay the ceiling rather than track usage.
the tier structure itself tells you what each company thinks its users are willing to do, calculate their exact needs, or just pay to stop thinking about it.
claude code adding AGENTS.md support feels like the ecosystem finally agreeing on one folder for telling coding agents how to behave.
the interesting part is letting ai work with real context without handing over the real identity behind it. that could make privacy way less annoying.
Introducing AgentCloak: Use any AI without sharing your real data.
Chinese AI services, ChatGPT, Claude, doesn't matter.
You probably try to hide details before asking: different names, fake numbers, no address.
But then the answer's useless because the AI is missing actual context.
AgentCloak runs in your browser.
It swaps your sensitive info for realistic fakes before sending anything, then swaps your real info back into the response.
You get what you need.
The AI gets nothing about you.
Works entirely in-browser.
Already trusted by some of the biggest companies in the world.
Now free to use.
agentcloak.ai/
someone asked why i don't just buy the shelf. cheaper, faster, no argument there. still can't give a better answer than "i wanted to make the shelf."
build a walled garden to lock users in. panel 2: build a slightly better walled garden. panel 3 (glowing): make your tool interoperable with two competitors. panel 4 (galaxy brain): win anyway because the product was just good.
antigravity cli ships cross-compatible with claude code and cursor skills, out of the box, per independent review coverage.
worth noticing that decision. google could have built a closed ecosystem to lock people into antigravity specifically. instead they made it interoperable with two direct competitors' plugin formats, a company confident in its own product doesn't need to trap people inside it to keep them.
interoperability isn't generosity here. it's what a company does when the product can actually win on its own merits.
spent longer matching grain on these two boards than i spent cutting them. nobody will ever ask about it. i'll know if i skipped it.
parallel threads that keep running after you leave turns claude code from a tool you sit with into something you can hand work to and check later.
Projects now run from one conversation, starting in Claude Code. You describe what needs doing, and Claude directs parallel threads that keep working after you close your laptop.
In beta today for select Pro and Max users in cloud sessions; coming to all Claude users soon.
curious how others handle this: when a vendor sunsets a cli tool with a 30-day window, do you migrate immediately or wait to see if the replacement matures first? asking because antigravity shipped without feature parity and i'm not sure waiting was wrong.
google announced gemini cli's shutdown at i/o on may 19, 2026, and stopped serving requests on june 18. thirty days between announcement and hard cutoff, with no grace period behind it.
worth noticing what that timeline communicates. a company confident in its migration path gives you six months. thirty days says the replacement wasn't ready and they needed people off the old thing anyway, antigravity cli launched without feature parity, which supports that read.
the shutdown date isn't the story. the gap between it and the announcement is.
this is where 3d generation gets really interesting. filling in what the camera can’t see while keeping the visible geometry intact is a big step.
Images rarely show an object’s full 3D geometry.
At #ECCV2026, our research team introduced Axolotl3D, a multimodal and occlusion-aware 3D generation model. It combines images, camera data and partial geometry to reconstruct missing regions while preserving observed ones, achieving state-of-the-art results across single- and multi-view settings.
Project page: research.nvidia.com/labs/sil…
openai launched chatgpt atlas as a standalone browser in october 2025. by july 2026, nine months later, they announced it was being discontinued entirely, folded into the regular chatgpt desktop app instead.
worth noticing what that reversal actually says. a standalone browser is a much bigger bet than a feature, its own install, its own market position against chrome. abandoning that after nine months means the standalone version wasn't earning its keep as its own product, even if the underlying capability was worth keeping.
the capability survived. the separate app it was supposed to justify didn't.
picked a book back up after leaving it half-finished for two months. remembered the plot faster than i expected. remembered why i stopped slower.
giving an ai agent permission to click, type, and act in my actual browser. it caught itself before a high-risk action and asked first. still checked twice.
what's harder for an ai company to raise money on?
a feature people demo well
or a claim about who's actually going to use it
replit raised $400m at a $9b valuation, tripling in six months, the same week they launched agent 4.
worth noticing the timing wasn't a coincidence. investors weren't betting on a feature list, they were betting on replit's claim that non-developers are becoming its most effective builders, people who know what to build better than how. that's a market bet, not a product bet.
the funding round is the real headline. agent 4 is just the proof they pointed to.
the /boost command might be the one i’d use most.
handing a messy developer task to a focused multi agent workflow sounds way better than trying to orchestrate everything yourself.
It's been a busy couple of weeks for the @antigravity team. Here's a few updates they've made ⬇️
> AlphaGenome Atlas Skills are launched to accelerate genomic discovery using AI agents.
> /boost command for tackling complex developer tasks through a focused multi-agent reasoning workflow.
> /teamwork-preview is improved and powering breakthroughs in mathematics, hardware simulation and open source optimization.
> Interactive Generative UI Artifacts creation is now available, ranging from data visualizations to 3D explanatory simulations that can be rendered seamlessly inline and in your artifacts panel.
> Updated Git Version Control System and Terminal, giving more flexibility to track changes to your repository and running tests, scripts, and commands directly in the embedded terminal.
players will forgive launch problems surprisingly fast when the developers own them and fix them.
silence is what turns server issues into a trust problem.
