🤖AI Newsletter
April 25, 2026 · 04:45 Uhr
1OpenAI Chief Scientist calls recent years of AI development "surprisingly slow"
THE DECODER OpenAI signals with GPT-5.5 and statements from its chief scientist that previous progress was merely preliminary stages and larger breakthroughs are imminent. This suggests sustained technological dominance and could pressure competitors while recalibrating investor expectations.
2GPT-5.5 leads AI rankings, costs 20 percent more and hallucinating more frequently than competitors
THE DECODER OpenAI regains market leadership with GPT-5.5 in AI benchmarks despite a 20% price increase, establishing a competitive advantage in the proprietary model segment. However, higher hallucination rates indicate quality tradeoffs that could impact enterprise adoption and customer trust, while the premium pricing strategy signals confidence but may accelerate competition from cost-efficient alternatives.
3After Andrulis' departure: Canadian startup Cohere acquires Aleph Alpha
THE DECODER Cohere acquires German AI startup Aleph Alpha following the founder's departure, with the Schwarz Group investing 600 million dollars. The acquisition signals consolidation in the European AI market and shows that national champions are failing against US competitors. Germany loses a beacon of hope in generative AI while Cohere strengthens its market position.
4Anthropic explains why Claude Code performed poorly for weeks
THE DECODER Anthropic resolved multi-week quality issues in Claude Code by identifying and fixing three error sources, but also signaled weaknesses in the QA process of a key product. The announced stricter controls aim to regain trust from enterprise customers and minimize competitive disadvantage versus OpenAI/GitHub Copilot.
5DeepSeek V4 is the largest open-weights model and significantly undercuts competitors on price
THE DECODER DeepSeek delivers massively scaled open-source AI models with V4-Pro/Flash that compete with proprietary competitors (OpenAI, Google, Anthropic) at significantly lower prices. This substantially increases margin pressure in the AI market and democratizes access to powerful models, putting established US providers under pressure and redistributing value chains.
Tokens: 1,388(834 in · 554 out)