{ "@type": "Dataset", "name": "@akibablade on X", "description": "Followers, posts and engagement for @akibablade.", "url": "https://tweetindex.com/account/akibablade", "isAccessibleForFree": true, "creator": { "@id": "https://tweetindex.com/#organization" }, "variableMeasured": [ { "@type": "PropertyValue", "name": "Followers", "value": 41033 }, { "@type": "PropertyValue", "name": "Following", "value": 2955 }, { "@type": "PropertyValue", "name": "Posts", "value": 6445 }, { "@type": "PropertyValue", "name": "Collected posts", "value": 7 }, { "@type": "PropertyValue", "name": "Views on collected posts", "value": 3711 }, { "@type": "PropertyValue", "name": "Likes on collected posts", "value": 21 } ], "distribution": [ { "@type": "DataDownload", "contentUrl": "https://tweetindex.com/api/accounts/akibablade", "encodingFormat": "application/json" } ], "@id": "https://tweetindex.com/account/akibablade#dataset" }

Akiba ✓

@akibablade · joined 31 Mar 2009

Not every story has 2 sides, some have 5, others only 1 | EIC @CryptoSlate | Techno-Progressive Decentralist Libertarian | Opinions my own

41 033Followers

2 955Following

6 445Posts total

3.7KViews on collected posts

โพสต์ล่าสุด

Akiba ✓@akibablade7 Aug

@akibablade @FrameworkPuter @AMD reasoning time isn’t free—on llama.cpp, a smaller quantized model with enough vram often beats a larger one that spends tokens searching. was ace measured at fixed token budgets? 3 views · 0 likes · 0 reposts · 0 replies Open on X →

Akiba ✓@akibablade7 Aug

Offline AI models don’t automatically perform better when given more reasoning time... I benchmarked these on my @FrameworkPuter @AMD Ryzen 395 Max+ 128GB Fast practical ACE score: 🥇 Qwen3.6-35B-A3B — 91.5 🥈 GLM-4.7-Flash — 88 🥉 Gemma 4 12B — 87 🎖️ GPT-OSS 20B — 82.5 🎖️ https:/

1.5K views · 4 likes · 0 reposts · 3 replies Open on X →

Akiba ✓@akibablade7 Aug

In my benchmark, extra reasoning improved Gemma 4, left GPT-OSS unchanged, and reduced the practical scores of Qwen3.6, GLM-4.7-Flash and BTL-4 Compact. More thinking is only useful when the model and runtime handle it well. Reasoning scores: Gemma 4 12B — 91 Qwen3.6-35B-A3B 275 views · 2 likes · 0 reposts · 0 replies Open on X →

Akiba ✓@akibablade6 Aug

https://t.co/i6elkFBgA8 629 views · 4 likes · 0 reposts · 1 replies Open on X →

Akiba ✓@akibablade3 Jun

don't sell your bitcoin:native ... that's not what I meant to say https://t.co/cjZZfIWBYk

2:08

480 views · 7 likes · 0 reposts · 4 replies Open on X →

Akiba ✓@akibablade2 Jun

Bitcoin is at the most important price level in recent history right now. Lose it and the floor becomes $48k and then $16k fast! Where's the bitcoin:native bid?! https://t.co/73wOvJT0rC

530 views · 2 likes · 0 reposts · 1 replies Open on X →

Akiba ✓@akibablade26 May

Same Shit Different Day The test lasted less than 24 hours and markets are about to open again after a long weekend. So far the Bitcoin proxy trade is unfazed so are bonds... let's see what happens when Wall Street gets online eh? Have markets given up on trading tweets yet?

307 views · 2 likes · 0 reposts · 0 replies Open on X →

Growth & engagement

How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.