Rish retweeted
Astra: stronger computer use and long multi-step work. Phased rollout, extra safety checks
Fable 5.1: better coding and 75% cheaper cache reads for agents that remember context
Gemini 3.8 Flash: fast coding workhorse, plus a gated Cyber variant
Qwen and Muse Spark: good enough at a price that makes loyalty look expensive
Rish retweeted
Everyone keeps saying better models will solve long-horizon tasks. Nvidia just proved the opposite. Claude Opus 5 alone: ~30%. Same model + solid agent harness: 100%. Stop waiting for the next model. Start fixing the system around the one you already have.
Rish retweeted
Do whatever it takes to find your real passion. was deep n crypto for years and it honestly ruined my life. But the last 3 months I’ve been surviving on 4 hours of sleep, skipping meals, & workouts I discovered Hermes and local AI's I finally found the thing worth going all in on
Rish retweeted
Anyone interested in collaborating on the design and development of Agenvoy’s WebUI?
I’m good at architecture and coding, but my UI aesthetic sense is really limited…
The current Web Dashboard is functional, but it’s still a long way from being both good-looking and easy to use.
@grok , introduce Agenvoy
Advice for everyone:
Set up an AI agent to run 24/7. It doesn’t even matter that much what it’s actually doing.
Just setting it up and running it is where the real learning happens.
And in your free time, scroll the AI tab on X.
Short-form brainrot on Reels and TikTok won’t add a single thing of value to your life.
Going forward, AI will be used in literally every field, and your competitive edge will be defined by how up-to-date you stay with the latest info.
Rish retweeted
Anthropic has reportedly overtaken OpenAI in revenue run rate and secondary market valuation for the first time, driven by strong demand for its coding and enterprise AI tools. This marks a significant shift in the competitive landscape between the two leading AI labs.
Rish retweeted
Hunt like crazy, experiment, build shit, fail daily, until you find that one thing that owns your mind 24/7
That’s when life actually starts feeling worth it
Rish retweeted
Stanford researchers just dropped OpenJarvis — an open-source framework for personal AI agents that run fully on your device by default.
From Hazy Research and Scaling Intelligence Lab at Stanford, part of their Intelligence Per Watt project.
It pairs cleanly with Ollama, gives you local memory for your docs, simple tools, and presets for stuff like morning briefings or deep research.
One quickstart script spins up the backend, browser UI, or desktop app.
It tracks real efficiency numbers (energy, latency, cost) and even improves from your own usage traces.
Cloud stays optional.
Big step for builders who want agents that actually stay private and under your control instead of living in someone else’s data center.
Blog: scalingintelligence.stanford…
👇GitHub link in comment
Rish retweeted
My advice to founders in 2026: spend tokens, not headcount.
Record everything. Make your company queryable. Build self-improving loops.
Before long, AI won’t just help you operate your company. It will make it self improving.
Don't think AI adoption, think AI transformation.
This is the biggest shift in how startups get built since cloud computing.
Rish retweeted
AI Agent Focus: Daily Maturation Loop
Run a "Failure Autopsy + Agent Self-Debug" session for 45–60 minutes.
How to do it (step-by-step):
Pick one recent agent run that didn’t perform as expected (or deliberately stress-test a new behaviour with 8–12 diverse inputs, including edge cases).
Log every failure mode with exact inputs, outputs, tool calls, and token usage.
Feed the entire log back into your agent (or a separate evaluator agent) with a structured prompt like: “You are an expert agent debugger. Here is a failed execution trace. Identify root causes (prompt ambiguity, tool limitations, state management, model drift, etc.), rank them by impact, and propose 3 concrete fixes with updated prompt/tool code snippets.”
Implement the top fix immediately and re-run the same test batch to measure improvement.
Why this is especially powerful for AI Agent builders:
Agents are complex systems with many failure surfaces (reasoning loops, tool integration, memory, evaluation). Random tinkering wastes days; systematic autopsies turn failures into rapid learning.
You train your own “meta-agent” skills at the same time you improve the product.
It creates natural content (screenshots of before/after traces, insights) for X later.
Edge cases covered: works whether you’re using LangGraph, CrewAI, AutoGen, custom Llama Index workflows, or raw OpenAI/Anthropic tool calling.
Expected outcome: You’ll usually find at least one high-leverage improvement (better error handling, clearer few-shot examples, missing tool, or state reset logic) that compounds over the week. Do this daily and your agents mature dramatically faster than “just keep building.”
Rish retweeted
Market conditions like this usually spawn the best products, and the best memes.
Ignore price, keep innovating. And post more memes.
Rish retweeted
Update on the meeting; according to Axios Defense Secretary Pete Hegseth gave Dario Amodei until Friday night to give the military unfettered access to Claude or face the consequences, which may even include invoking the Defense Production Act to force the training of a WarClaude
if this market has you doubtful, stop worrying immediately
you already possess the most valuable resource in the world: time
no matter if we chop, drop or top, the trenches will still be here tomorrow, and the day after, and so on
before you thrive, you must first survive.
Rish retweeted
PNL OR ZERO
Pure trading is dangerous - because you're only paid for results. Max results
In other endeavours, you get rewarded for participation and skill acquisition - because there are avenues to earn with your skill while structure & capital is provided by others - i.e. a job, content creation, lecturing, etc.
Example:
A software engineer will be paid if he can display skill, he doesn't have to build a working product to earn. He can work on a section of the product and still be rewarded handsomely.
Trading isn’t like that.
In trading, you’re paid nothing unless you win. ABSOLUTELY NOTHING.
Your skills can be truly great. Narrative recognition, wallet tracking, TA etc. But your only hindrance could be your emotions. And because of that, you’re seen as not good enough - or worse, you view yourself as a failure - just because the results (profits) aren’t there yet.
To make it worse, your success can be erased in seconds - a single liquidation event (10-10-25 is an example), a black swan, or a small tilt episode can wipe out weeks or months of good trading.
The above observations makes trading, as a full-time, all-in career path, extremely dangerous to your longevity and sustainability - especially if you haven’t secured “f*ck you money” (let’s say mid-7 figures or 8).
These are thoughts I have pondered on and even written an article about but it is in your interest that I double down for your understanding.
What is the point?
My point extends over a longer timeframe - five years or more - for those of you planning to stick around for the long haul.
If your earning capacity in crypto comes solely from trading, especially after spending time developing the skills to navigate the blockchain, then you’re setting yourself up for existential failure. One hack. One error. One emotional spiral. And you're back to zero. No plan B.
What is the counter?
I love trading but I have thought of my participation in Crypto over a longer timeframe.
The ultimate goal is to get paid for my process & skills in crypto- and trading becomes an avenue to multiply these funds.
I have two degrees that correlate strongly to crypto. I do not plan to let them go to waste.
Some of you are highly talented - artists, communicators, teachers, motion graphic designers, software engineers etc but you have never thought of a way to integrate these skills into your crypto journey.
This is your cue to map out a longer timeframe plan.
CREATE A CRYPTO RESUME BY ALL MEANS POSSIBLE.
There are jobs in crypto, there are also jobs you can create for yourself in crypto.
Its a longer timeframe plan so you don't need to rush it. But you need to display it while you trade up your net-worth.
| You're an artist? Make art. Post it. Offer to create for a project - even for free.
| You can explain things well? Make educational content - written or visual.
| You can code? Build something. Make tools. Open source them.
| You're philosophical? Post your ideas. Start writing.
| You’re naturally curious or interesting? Start a podcast. Doesn’t have to be weekly - just publish something.
| You're funny? Shitpost.
| You're a finance guy? Break down tokenomics. Simplify complex things.
Value attracts financial reward - especially when you’re not actively chasing it.
Having zero evidence of your skills outside of trading is dangerous. There are many ways to get paid in crypto, but it starts with providing value.
Also, you can just apply for jobs. That experience will compound and set you up for even bigger six-figure roles down the line.
Your trading layer needs to be solid regardless as trading is the ultimate skill - but it’s for multiplication, not survival.
Set up for the future. HIGHER TIMEFRAMES.
Build value beyond your PNL.
Love, Waller.






















