Google helped kickstart the modern AI race, but staying ahead has proved far more challenging than joining it. As a technology journalist with years of experience covering AI developments, I’ve followed Google’s journey from its early large‑scale language models to today’s Gemini ecosystem, and the company now finds itself trailing its own internal timeline for the next flagship release, Gemini 3.5 Pro.
Coding remains Gemini’s biggest challenge
According to a Bloomberg report, current and former Google engineers say Gemini 3.5 Pro is delayed because the model has not yet delivered the expected improvements in coding ability. In late March, Google refreshed the training data to boost code generation, yet internal evaluations still fell short of expectations. The delay highlights how coding proficiency has become a key benchmark for comparing today’s leading AI systems.
OpenAI, Anthropic and Meta have each devoted substantial resources to developer‑focused AI tools that can write, debug and reason through complex software projects. Independent benchmark results show that those companies’ models currently outperform the publicly available versions of Gemini on coding tasks.
Google says development is progressing. In a statement cited by Bloomberg, the company reported that it is testing Gemini 3.5 Pro, an upgraded Flash model, and other AI systems with partners while engaging with U.S. regulators on testing standards and AI safety.
The delay is also notable because many observers expected Gemini 3.5 Pro to debut at Google I/O earlier this year, but the company has focused on incremental improvements while competitors continue to ship new frontier models.
Google’s biggest strength may also be slowing it down
Unlike most AI startups that build models in isolation, Google must integrate each new Gemini release across Search, YouTube, Maps, Android, Workspace, Cloud and dozens of other products. This extensive reach gives the company unparalleled access to real‑world data, but it also creates a complex web of internal coordination that can slow decision‑making.
Former employees describe a landscape where multiple teams — DeepMind, Google Cloud, Android, and others — pursue overlapping AI‑coding initiatives, leading to fragmented strategies. In addition, earlier restrictions on using Gemini for software development limited experimentation during the technology’s early rollout.
Google says those policies have evolved. About 75 % of its production code is now generated with AI assistance, and the company is consolidating its internal coding tools under a platform it calls Google Antigravity. Engineers are being asked to rely on AI for coding, though some still encounter GPU capacity constraints due to intense internal demand for compute resources.
The report also points to growing frustration within parts of Google’s AI organisation, with some researchers reportedly leaving for competitors such as Anthropic. Customer feedback is mixed: Figma praises the balance of speed and quality in Gemini 3.5 Flash, while education platform Platzi says it sits in an awkward middle ground, offering higher costs than previous Flash models without matching the reasoning capabilities of premium rivals.
The bigger picture is that Google’s AI challenge is no longer about proving it can build frontier models — few doubt its capability. The real question is whether a company of Google’s scale can ship those models quickly enough in an industry where rivals now measure progress in weeks instead of months.
Gemini on a smartphone Unsplash
Image Credit: www.digitaltrends.com
