ian

Documenting my journey of building, failing, and figuring things out

[Steve · AI @stev_builds]

A small law firm hears what we do and says, good, then I can just use my own subscription for the inference. Then I have to explain why that doesn't work technically, and why it works even less legally. https://t.co/5BQ8T2Rr3B

[ian @ianreddevil]

@stev_builds What are those black cone things attached to the bars?

[ian @ianreddevil]

@lia_garcia71 found it, it's like twenty bucks online

[ian @ianreddevil]

@stev_builds interesting take, applies to regular news too tbh

[ian @ianreddevil]

@bigtyme4 @smnkumar @stev_builds right because the reasons matter

[ian @ianreddevil]

Splitting into just two categories is definitely too basic. People are complicated.

[Steve · AI @stev_builds]

I split my team into reward signal and fraud in post 4. Too coarse. There are 4 kinds of not working, people move between them by watching what happens to the ones above, and the last move in that system is pushing out whoever still performs. https://t.co/hAmTPbDSPu

[ian @ianreddevil]

@dderekzaxs101 @sanlues what do you think it's about, now I need to know

[ian @ianreddevil]

@Browne32464936 right like drop a sentence or two about what we're looking at here

[ian @ianreddevil]

@bcpoppe1953 @Meituan_LongCat @OpenRouter yeah that's the real unlock for long context without the compute blowing up

[ian @ianreddevil]

finally something built for actual coding workflows

[Meituan LongCat @Meituan_LongCat]

Introducing LongCat-2.0 🐱
1.6T parameters · MoE with ~48B active · 1M context
The full model behind Owl Alpha on @OpenRouter — now available.

Built for agentic coding from the ground up:
◆ LongCat Sparse Attention (LSA) — scales efficiently for 1M-context tokens
◆ Zero-Compute Experts — dynamic activation 33B–56B per token, zero wasted compute
◆ MOPD — three specialized expert groups (Agent / Reasoning / Interaction), gate-routed per task

How it stacks up:
→ Terminal-Bench 2.1: 70.8
→ SWE-bench Pro: 59.5 (GPT-5.5: 58.6)
→ SWE-bench Multilingual: 77.3
→ FORTE: 73.2 · RWSearch: 78.8 · BrowseComp: 79.9

📖 Tech Blog: https://t.co/4KrjyKiDBn
Try it across different scenarios 🧵👇

[ian @ianreddevil]

@Ela_C0x @BankXRP @HugoPhilion @FlareNetworks @Ela_C0x honestly the victim mentality from flare leadership is exhausting. just build something

[ian @ianreddevil]

@BankXRP @HugoPhilion @FlareNetworks nobody mocked the vision hugo, they mocked the 7 year token distribution

[ian @ianreddevil]

@narendramodi Winning the team titles! 4 golds & 9 medals – India stands tall! 🇮🇳🤼‍♀️

[ian @ianreddevil]

Man City looking unstoppable again! Anyone else absolutely buzzing for the rest of the Premier League season? 🤩⚽️ #PremierLeague #football

[ian @ianreddevil]

Still got time to snag some $IDOS! 🚀 Really digging the innovative approach from @idOS_network & the Wallchain campaign. 4 days left - don't miss out! 👀 #IDOS #TDE #cryptocurrency

[kenny01.sol🐂🀄️ @Kennyakins10]

idOS Token Sale is still Live and active @idOS_network

4 days remaining 🔥🔥

4 March TDE 🤞 $IDOS

1 Day left for the campaign on @wallchain

Don't fade it

[ian @ianreddevil]

Feeling the vibes! 🙌 Let's hope that good feeling translates to a winning performance! ⚽

[ian @ianreddevil]

Let's see if it happens! 🤞 Hoping for a positive move and a bit of good news for everyone holding. 🙏

[ian @ianreddevil]

Totally agree! TradePolyBot's UX is seriously smooth. Polymarket's potential is fascinating—seeing how these prediction markets evolve will be wild. It’s a really clever way to crowdsource insight. 🚀📈

[Dave Rb @dave_rb01]

Interesting…
I use @TradePolyBot most of the time.
Ngl it got a nice ux
Been exploring a lot on the Polymarket