So many games out there that are excellent and theyre like Pokémon but they’re not pokemon and the reason I don’t want to play them is simply because they’re not Pokémon.
I’ve been sponsored by amazing brands like:
Replit,
Chatbase,
Apollo,
Poly,
Airops,
Emergent,
Hostinger,
Higgsfield,
Omnisend,
Livedocs,
Funnelfox,
Superapp,
Articos,
Viktor,
Wircle,
Okara,
Gamma,
Icon,
Arcads, etc.
All because I kept posting online and sharing declarative (practical/actionable) knowledge.
You should do the same!
God damn. I cannot recommend more strongly that everyone, EVERYONE, whether you’re an Eng, pm, or even investor/finance bro, watch this. At LEAST the first half. @poteto was on the react compiler team at meta and an EM at Netflix.
What struck me most is the CLEAR skepticism she once had about agents, how she talks through all the problems IN DETAIL, but how she systematized solving those problems and sort of climbed up the hill of agent incompetence despair.
2 weeks ago I launched my first game Luminary
Week 1 grossed $100k+ in sales. Only 2% of Steam games do that, and with No Funding, No Publisher, No Ads, just hard work.
Stats Today
Net: $100k
After Steam's cut: $70k
All Reviews: Mostly Positive
English Reviews: Very Positive
the best tools are the ones you build.
I completely revamped https://t.co/mDYZvvxTif and now i'm using it DAILY
-> it's becoming the sharpest market research tool I have ever had
I connected it to TONS of sources, top X signals, filings, news, web search, reports, anything i can find to enrich my thesis around a topic or a company.
- it's the absolute best intelligent layer to catch signals on the market im(h)o
- it also has daily briefs and actionable setups
- you can use it to explore evidence backed theses, bull/bear cases, risks and upcoming events.
I don’t need more news, i need to know what the news changes:
- does it strengthen the thesis?
- break it?
- move a price target?
- create a setup?
That is pretty much the system I’ve been building
Have a go, all feedback is welcome!
IBM's current CEO Arvind Krishna (from India) became CEO in April 2020.
He immediately started firing white employees. He increased the H1B imports from India, plus he shifted massive numbers of jobs to India.
IBM almost immediately lost the US patent 1st place position.
i finally cracked frontend design with AI... without using any skill
frontend needs taste to not look like slop, and i'm more of an engineer than an artist, so i'm genuinely bad at it
my way around it: collecting inspirations and modules from other people, building a lego of components
resources you can send to your agents:
- https://t.co/7CUfh1jSmr
- https://t.co/hKiKiyjLd8
- https://t.co/XMkWlWenQv
- https://t.co/k3p8MhfvCj
- https://t.co/PTkhqaLgU7
it's VERY easy for agents to edit components once they have the code
you send the link, your agent fetches the full list of components, then you ask it to find the best way to integrate them into your frontend foundation
the easiest way i found to make your UI/UX drastically better without being a genius designer
this is pure f*cking treasure
YOUR REVIEWER AGENT AGREES WITH YOUR CODER FOR NO REASON
two researchers measured it, named it, and fixed it with three lines.
> https://t.co/gIYYKcLV4C
the standard cure for weak agentic review is more agents. it doesn't work: you scale the count, returns drop, and the reviewers start agreeing with each other without evidence. the paper calls it false consensus, and here's how it kills a real bug: a reviewer writes a confident-sounding objection, the critic backs down, the flag gets dropped, the bug ships to prod.
the cure isn't more agents. the critic is forced to end with exactly one of three verdicts:
AGREE
DISAGREE_EVIDENCE: <code citation>
DISAGREE_CONCERN: <epistemic objection>
that's the whole trick. "i disagree" now has to arrive with a line of code that refutes the flag - or say out loud that it has nothing to show.
what three agents did to five:
→ LiveCodeBench: 87% pass rate, +5 pp over the five-agent setup, with 2 reviewer roles against 4
→ on the hard split: 43 of 57 against 39
→ SWE-bench Verified: 75.2% vs 72.6% for the five-agent setup and 71.6% zero-shot, N=500
→ SWE-PRBench: F1 0.533, best in the split - the naive version sat at 0.457 until disagreement was made explicit
→ the honest cost they report themselves: roughly 4.5x more tokens than zero-shot
no code shipped with the paper. it's two prompts and three verdict lines. both prompts verbatim in the reply.
take it 👇
#1 underrated skill for your harness?
Ponytail
- It forces your agent to think like the laziest senior dev
- It can save you HOURS of research and work
It's like the developer who looks at 50 lines… says nothing… and replaces them with one.
Works with Claude Code, Codex, Hermes Agent, Copilot CLI, Pi, OMP, OpenCode, Cursor, Gemini, and more.
https://t.co/5CxRPYLsI3
Introducing Higgsfield in Blender.
> Prompt the scene and build the blockout
> Describe the camera move and get it animated
> Adjust anything by hand
> Reblock the whole shot in seconds
Now available via Higgsfield MCP or Supercomputer.
I finally got the water to interact with the environment. The hose wraps around trees, the stream stops where it hits, and it pushes things around.
Join our Steam Playtest now!
Mac Studio M5 Ultra 256gb vs 2xDGX Spark
Price : $10,799 vs $10,000
Prefill : almost tie (expected)
Decode : 1,200GB/s vs 273GB/s
Mac Studio is definitely more worth the price.