Every post of theirs in the library
Loading their posts
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)
Writes about AI and Engineering. 82% of posts open with an observation. A typical post reaches 339 likes.
- Posts
- 33
- Likes
- 16K
- Typical
- 339
- Best
- 2.5K
What they write about
How they open
Every post of theirs in the library
| Post | Topic | Likes | Replies | Reposts | Score | Posted | Actions |
|---|---|---|---|---|---|---|---|
| @teortaxesTexI really admire them. I respect the projects of other open teams too, but NOBODY comes clo… | SaaS | 2.5K | 24 | — | 52 | ||
| @teortaxesTex> $5.5M for Sonnet tier it's unsurprising that they're proud of it, but it sure feels like… | AI | 1.4K | 45 | — | 44 | ||
| @teortaxesTexI have tried to use OpenRouter Fusion API with cheap open models only, and saw reasoning t… | Engineering | 1K | 49 | — | 45 | ||
| @teortaxesTexDeepSeek: «We plan to raise the overall pricing for DeepSeek API services in the near futu… | Revenue | 857 | 62 | — | 48 | ||
| @teortaxesTexlistening now Xi makes a speech that's basically DeepSeek recruiting copy. AI as steam and… | AI | 728 | 20 | — | 42 | ||
| @teortaxesTex> All benchmark results are reported under INT4 precision. Do you understand what a flex t… | AI | 717 | 21 | — | 42 | ||
| @teortaxesTexJunyang had been announcing Qwen models for… not so long, 2 years? Yet it was more models … | AI | 669 | 18 | — | 41 | ||
| @teortaxesTexVP of Science at OpenAI hyped up that GPT-5 «found solutions» to 10 open «Erdös» (actually… | AI | 635 | 11 | — | 41 | ||
| @teortaxesTexafter giving Manus a spin I conclude it's a product devilishly optimized for influencers, … | AI | 613 | 26 | — | 38 | ||
| @teortaxesTexDeepSeek-V4-Flash is updated. «DSBench-FullStack is an internal full-stack development tes… | AI | 604 | 40 | — | 41 | ||
| @teortaxesTexMassively unexpected update from DeepSeek: a powerful, high-compression MoE OCR model. > I… | AI | 576 | 17 | — | 40 | ||
| @teortaxesTexGLM 5.2 is apparently the strongest Chinese model, and it'll be open this is quite a jump … | AI | 502 | 17 | — | 39 | ||
| @teortaxesTexanon do you realize that V4-Pro is straight up the strongest pretrained model we have? Lik… | AI | 467 | 16 | — | 38 | ||
| @teortaxesTex…It wasn't an intern's joke MiMo 2.5 (not Pro): > Trained on a total of ~48T tokens using … | AI | 429 | 10 | — | 37 | ||
| @teortaxesTexWestern open source frontier is back. This is on par with the largest of known Chinese pre… | Engineering | 397 | 10 | — | 37 | ||
| @teortaxesTexDo you get what this means anon [nerfed] GLM 5.2 can materially help in defending against … | AI | 345 | 15 | — | 36 | ||
| @teortaxesTexon the contrary, Huggingface should freak out about a world where OpenAI and Anthropic can… | AI | 339 | 12 | — | 36 | ||
| @teortaxesTexThis is important too Two more notes: - they don’t call models “exp” or “preview” as a jok… | AI | 319 | 12 | — | 38 | ||
| @teortaxesTex> built through continual pretraining on approximately 15 trillion mixed visual and text t… | AI | 311 | 6 | — | 34 | ||
| @teortaxesTexso what's the justification for openly releasing Opus 5 if it's de facto the strongest Ant… | AI | 266 | 21 | — | 34 | ||
| @teortaxesTexthe power of scaled RL + distillation (but big boy distillation, not cringe SFT on claude … | AI | 251 | 20 | — | 34 | ||
| @teortaxesTexAPI is overpriced vs subscriptions by 40-70x Do you know what this means? They have DeepSe… | Engineering | 249 | 14 | — | 33 | ||
| @teortaxesTexGrok seems to get tired and start lazily cheating later into the session (247K/500K), or I… | AI | 238 | 7 | — | 32 | ||
| @teortaxesTexThis is also what @ylecun means when saying that GPT-4 or "GPT-5000" doesn't have a cat's … | AI | 223 | 21 | — | 31 | ||
| @teortaxesTexRemember: @deanwball doesn't write USG policy anymore, but he has a fairly good idea of ho… | AI | 195 | 14 | — | 32 | ||
| @teortaxesTexOh well. America Winning once again. Wypipo innovate, the Chinese copy. I feared that it'l… | Marketing | 192 | 25 | — | 33 | ||
| @teortaxesTexOk these are pretty good scores …though some crucial agentic ones (NL2Repo, DeepSWE, Agent… | AI | 188 | 14 | — | 31 | ||
| @teortaxesTexDefeated with cheap rewriting, definitely defeated if you train a rewriter against a recon… | Revenue | 187 | 10 | — | 33 | ||
| @teortaxesTexSo speaking of benchmarks, what can be said of the new open Qwen? First, it completely des… | AI | 176 | 9 | — | 30 | ||
| @teortaxesTex> The depth of the computation graph for our present frontier models, including Astra, is … | AI | 171 | 7 | — | 34 | ||
| @teortaxesTexI'd actually say that "open source" as a dumping strategy makes little sense The point of … | Engineering | 164 | 10 | — | 30 | ||
| @teortaxesTex> We propose IndexShare, which reuses the same indexer across every four sparse attention … | AI | 152 | 8 | — | 29 | ||
| @teortaxesTexDeepSeek is trolling They don't say *what* they trained on, only "eh MXFP4" Hawks will now… | Science | 151 | 4 | — | 28 |