OpenAI 'Spud' pre‑training done
OpenAI’s next‑generation model, codenamed 'Spud,' has completed pre‑training and is being positioned to accelerate both research and commercial applications, according to early coverage. That pre‑training milestone suggests an imminent phase of evaluation and product integration for researchers and partners. (mindstudio.ai)
OpenAI reorganized product and oversight lines this week, renaming Fidji Simo’s product group to “AGI Deployment” and moving safety and security reporting to other senior leaders so Sam Altman can focus on fundraising and datacenter expansion. (theinformation.com) The company is winding down its short‑form video app Sora and its related Disney licensing arrangements to free engineering capacity and budget for higher‑priority projects. (bloomberg.com) Internal reporting places the milestone that triggered those moves in late March 2026, with senior leadership telling staff they expect a “very strong model” to be ready for downstream evaluation within weeks. (theinformation.com) Operationally, OpenAI is shifting compute and hiring toward large‑scale infrastructure: the company has public listings for Stargate datacenter roles and infrastructure data‑science positions that explicitly support deploying “the right compute… at the right time and place.” (datacenterdynamics.com) Hiring signals inside OpenAI’s research listings emphasize both traditional research credentials and production‑grade skills—job descriptions ask for first‑author publications or equivalent research output alongside experience building “high‑performance implementations” that work at large scale. (openai.com) For candidates weighing routes into elite labs, public hiring pages at DeepMind and OpenAI show the dual premium: a PhD and strong publication record remain explicit advantages at DeepMind, while corporate research roles increasingly list system and infra engineering skills as core requirements, reflecting the compute‑heavy, deployment focus that Spud’s timetable and OpenAI’s datacenter push make concrete. (job-boards.greenhouse.io) The longer trend shows talent flowing to industry labs: studies and industry analyses document a steady migration of top AI PhDs into corporate research over the past decade, with industry placements rising markedly (a cited 2019 estimate put post‑PhD industry placement around 65% versus ~44% in 2010), underscoring why candidates targeting model‑deployment work should prioritize practical scaling and systems experience in addition to ML theory. (link.springer.com)
