PrimeIntellect

23 atoms · 跨 13 天 · 首见 2026-04-29 · 最近 2026-06-22

三色: 🟦 fact 16 · 🟥 take 7 · stance ▲7/▼0/◆5

来源: X 23

时态: fresh:21 · aging:1 · stale:1

标签: 好观点:8 · 好数字:7 · 好思考:6 · 好信源:6 · 好问题:2

叙事 (narrative) (6)

• 2026-06-01 · 角色
  Proud to be part of the coalition
  tags: 好观点·好思考 · → daily
• 2026-05-14 · AI 研究自动化
  We started automating AI research on nanogpt-speedruns & achieved new records
  tags: 好观点·好思考 · → daily
  value: direction=started automating
• 2026-04-29 [老化中] · 训练集群规模
  3k h200 is a lot here (compared to dsv3 for instance)
  tags: 好思考 · → daily
  value: qty=3k h200 · type=GPU
• 2026-06-18 · 硬件提供
  Special shout out to our hardware provider @PrimeIntellect . this model is running on gb300NVL72s and the have been absolutely amazing. could not recommend getting your large scale hardware from them highly enough
  tags: 好观点 · → daily
• 2026-06-03 · 提供 hosted training 服务 50% 折扣
  Sign up to https://t.co/DkOji4KKq3 with promo code "Ziems" to get 50% off
  tags: 好观点 · → daily
  value: qty=50% off · date=ongoing
• 2026-05-20 · 声明 reward hacking 是 RL 最难问题
  Reward hacking is the hardest problem in RL.
  tags: 好观点 · → daily

立场 (position) (1)

展开 1 条老旧 / 已过期

• 2026-05-20 [老旧] · Reward hacking 是 RL 最难问题
  Reward hacking is the hardest problem in RL.
  tags: 好观点 · → daily

事实 (fact) (16)

• 2026-06-22 · achieved GLM-5 on agentic SWE tasks at 131k context and sub-5-minute step time
  The result: GLM-5 on agentic SWE tasks at 131k context and sub-5-minute step time.
  tags: 好数字·好信源 · → daily
• 2026-05-31 · 用户请求对评估运行进行粗略代币估算
  rough estimate on how many tokens it takes to run an eval for x model would be good (even though error bars would be big).
  tags: 好问题·好思考 · → daily
• 2026-05-30 · 发布托管评估平台
  Today, we are launching Hosted Evaluations on the platform. Running evals is an infra problem: harnesses, sandboxes, hours of compute, hundreds of parallel runs. Running evals is hard. Until now.
  tags: 好观点·好思考 · → daily
  value: date=2026-05-30
• 2026-05-14 · GPT 5.5 和 Opus 4.7 迭代优化
  for 2 weeks GPT 5.5 and Opus 4.7 iterated on novel optimizations 10k runs & 14k H200 hours
  tags: 好数字·好信源 · → daily
  value: date=2 weeks · qty=10k runs & 14k H200 hours
• 2026-06-22 · releasing prime-rl v0.6.0
  Today we're releasing prime-rl v0.6.0 — enabling RL at trillion-parameter MoE scale on agentic workloads at the highest efficiency.
  tags: 好信源 · → daily
• 2026-06-22 · prime-rl is fully open source
  prime-rl is fully open source, and we're hiring systems engineers to take it further.
  tags: 好信源 · → daily
• 2026-06-04 · 加入NVIDIA Nemotron Coalition
  We're excited to join the NVIDIA Nemotron Coalition
  tags: 好信源 · → daily
• 2026-06-02 · 应用研究方向实习生招聘
  参加 PrimeIntellect 应用研究方向暑期实习
  tags: 好信源 · → daily
• 2026-06-02 · 招聘岗位数量
  _Join us at @PrimeIntellect to build the open stack for self-improving agents

Engineering

• MTS – Full Stack Software Engineer — SF/Remote, Full time

• MTS – GPU Infrastructure — SF/Remote, Full time, Hybrid

• MTS – Inference — Remote/SF, Full time, Hybrid

• MTS – Sandbox Platform — SF, Full time, On-site

• MTS – Security — SF, Full time, On-site

• MTS – Training Platform — SF, Full time, On-site

Research

• Research Engineer – Distributed Training — SF/Remote, Full time

• Research Engineer – Reinforcement Learning — SF/Remote, Full time

• Research Engineer – RL Infrastructure — SF/Remote, Full time

• AI Research Resident – Open Source AGI — Remote, Part time

Applied Research

• Evals & Data — SF, Full time, Hybrid

• Forward-Deployed — SF, Full time, On-site

• RL & Agents — SF, Full time

Compute / Finance

• Head of Compute — SF, Full time

• Strategy and Finance Lead, Compute — SF, Full time

Finance / Operations

• Business Operations Lead — Remote, Full time

• Chief of Staff — SF, Full time

• Founder's Associate, Business Operations — SF, Full time

Growth

• Account Executive — SF, Full time

• Head of Enterprise — SF, Full time

• Head of Growth — SF, Full time

• Head of Marketing — SF, Full time

• Revenue Operations Lead, AI Infrastructure — Remote, Full time

• Solutions Architect – AI Infrastructure — SF, Full time

• Technical Account Manager – AI Infrastructure — SF, Full time

Legal

• Head of Legal — SF, Full time

Others

• Internship — SF, Full time

• Open Application for Unconventional Talent — SF/Remote, Full time_

tags: 好数字 · → daily

value: qty=26 · date=2026-06-03

• 2026-06-02 · 新人入职状态
  need to update the pic with you and everyone else that just joined 🙌
  tags: 好思考 · → daily
  value: date=2026-06-03
• 2026-05-15 · idle compute as percentage of annual total
  less than 0.01% of our annual total - majority is utilized by customers
  tags: 好数字 · → daily
  value: qty=less than 0.01%
• 2026-05-14 · idle compute usage for AI research runs
  We let Claude Code (Opus 4.7) and Codex (GPT 5.5) run autonomously on the nanoGPT speedrun optimizer track using our idle compute. ~10k runs, ~14k H200 hours
  tags: 好数字 · → daily
  value: qty=~10k runs, ~14k H200 hours
• 2026-05-14 · Opus 4.7 创纪录
  Opus now holds the record at 2930 steps
  tags: 好数字 · → daily
  value: qty=2930 steps
• 2026-05-14 · 代理模型超越人类基准
  both agents beat the human baseline
  tags: 好数字 · → daily
• 2026-05-30 · 用户请求运行价格估算
  can we get a price estimate on runs
  tags: 好问题 · → daily
• 2026-05-21 · 基础设施管理
  our observability agent has been running since we launched hosted training - managing heterogeneous hardware from h100 clusters to nvl72 deployments wouldn't be possible without it. auto-detects infrastructure issues, references historical patterns from sqlite, and creates detailed github issues + slack alerts like this one
  tags: 好观点 · → daily

时间轴 (近 20)

• 2026-06-22 · fact · releasing prime-rl v0.6.0 · → daily
• 2026-06-22 · fact · achieved GLM-5 on agentic SWE tasks at 131k context and sub- · → daily
• 2026-06-22 · fact · prime-rl is fully open source · → daily
• 2026-06-18 · narrative · 硬件提供 · → daily
• 2026-06-04 · fact · 加入NVIDIA Nemotron Coalition · → daily
• 2026-06-03 · narrative · 提供 hosted training 服务 50% 折扣 · → daily
• 2026-06-02 · fact · 应用研究方向实习生招聘 · → daily
• 2026-06-02 · fact · 招聘岗位数量 · → daily
• 2026-06-02 · fact · 新人入职状态 · → daily
• 2026-06-01 · narrative · 角色 · → daily
• 2026-05-31 · fact · 用户请求对评估运行进行粗略代币估算 · → daily
• 2026-05-30 · fact · 发布托管评估平台 · → daily
• 2026-05-30 · fact · 用户请求运行价格估算 · → daily
• 2026-05-21 · fact · 基础设施管理 · → daily
• 2026-05-20 · position · Reward hacking 是 RL 最难问题 · → daily
• 2026-05-20 · narrative · 声明 reward hacking 是 RL 最难问题 · → daily
• 2026-05-15 · fact · idle compute as percentage of annual total · → daily
• 2026-05-14 · fact · idle compute usage for AI research runs · → daily
• 2026-05-14 · fact · GPT 5.5 和 Opus 4.7 迭代优化 · → daily
• 2026-05-14 · fact · Opus 4.7 创纪录 · → daily

---

← 实体目录 · 系统日志