OpenAI harness fix, EU AI gigafactories, and Cognition's SWE-1.7: AI roundup Aug 1, 2026

·7 views

OpenAI revealed on July 29 that GPT-5.6 Sol's poor ARC-AGI-3 benchmark score of 7.8% was caused by a flawed evaluation harness that discarded reasoning between moves, not by the model itself. Enabling retained reasoning and context compaction in the Responses API lifted Sol's score from 13.3% to 38.3% and allowed it to clear all six benchmark levels, a feat no frontier model achieves on the official leaderboard. The finding highlights that benchmark results reflect the combined effect of model, harness, and settings rather than model capability alone. Meanwhile, the European Commission launched a formal tender on July 30 to build up to seven AI gigafactories across the EU, backed by up to €10 billion in public funding and a €30 billion total target including private investment, with awards expected in early 2027. Separately, AI coding lab Cognition released SWE-1.7 in July, a new frontier coding model built on a Kimi K2.7 base and trained via reinforcement learning, featuring innovations in sampling stability, multi-continent training, and extended task horizons.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

How to Fairly Split Shared Kubernetes Costs Across Engineering Teams

Managing cost attribution in shared infrastructure like Kubernetes clusters is a growing challenge for engineering organizations, especially those with more than 50 engineers. Dr. Samson Tanimawo outlines a four-level framework ranging from no attribution to full charge-back, where team budgets are actually debited for their cloud usage. For shared services used across teams, costs can be distributed by equal split, usage-based allocation, or absorbed centrally, depending on the service's share of total spend. Untagged resources — often around 20% of total spend — should be assigned to a platform team to incentivize proper tagging discipline. Experience shows that after three months of cost visibility reporting, most teams voluntarily reduce waste by 15 to 25 percent.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How your robots.txt file silently blocks AI tools from ever citing your site

A website's robots.txt file determines whether AI platforms like ChatGPT, Claude, and Perplexity can cite it in answers, yet most sites configure this file without fully understanding the consequences. There is a key distinction between training crawlers, such as GPTBot and ClaudeBot, and retrieval agents like OAI-SearchBot and PerplexityBot — blocking the latter permanently removes a site from AI-generated citations. A 2024–2025 wave of blanket AI bot blocking led many publishers to inadvertently restrict both types, with audits finding major outlets including the New York Times, Reuters, and Wired allowing fewer than three of seven retrieval agents. CDN providers like Cloudflare can also silently inject rules into the served robots.txt file that contradict or override a site owner's own settings, sometimes even adding legal declarations the owner never authored. Experts recommend running a simple curl check on your live robots.txt to identify unintended blocks and ensure retrieval agents are explicitly permitted if AI citation visibility is a goal.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How to Fairly Split Shared Kubernetes Costs Across Engineering Teams

Managing cost attribution in shared cloud infrastructure like Kubernetes clusters is a common challenge for growing engineering organizations, according to a guide by Dr. Samson Tanimawo. The framework outlines four maturity levels, from no attribution at small scale to full charge-back systems where team budgets are actually debited. Visibility-first approaches, such as monthly show-back reports, can prompt teams to voluntarily cut 15–25% of wasteful spending within three months. Special cases like shared services, burst workloads, and untagged resources each require specific handling strategies to avoid internal conflict. Assigning untagged cloud spend to a platform team is recommended as a fast way to enforce proper resource tagging across the organization.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Why a Portfolio Matters for Every Tech Professional, Not Just Designers

A professional portfolio is increasingly considered essential for all technology roles, including developers, data scientists, product managers, and QA engineers, not just designers. While resumes and cover letters provide a brief snapshot of qualifications, they rarely have space to explain the reasoning, challenges, and impact behind a candidate's work. A portfolio fills that gap by showcasing problem-solving approaches, decision-making processes, and measurable outcomes in depth. Hosted online, it can be discovered through platforms like LinkedIn, GitHub, or search engines, creating a lasting professional impression. It can include case studies, blog posts, open-source contributions, certifications, and speaking engagements, effectively serving as a personal brand beyond the traditional job application.

0 comments Read more at DEV Community