SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World
Back to feed
Programming

OpenAI Upgrades Auto-Review to GPT-5.6 Luna to Cut Agentic Workflow Costs

0
·6 views

OpenAI has replaced GPT-5.4 with GPT-5.6 Luna as the model powering Auto-review in the ChatGPT app and Codex CLI, targeting cost reduction in automated agentic workflows. Luna is the fastest and most affordable tier in the GPT-5.6 family, priced at $0.20 per million input tokens and $1.20 per million output tokens following an approximately 80% price cut announced on July 30, 2026. Auto-review is a safety mechanism that automatically approves low-risk requests in Codex, aiming to reduce manual intervention without removing oversight entirely. OpenAI has suggested the upgrade could make agentic workflows up to ten times cheaper, though actual savings will vary based on token volumes, workflow design, and how often tasks escalate to higher-capability model tiers. The GPT-5.6 family also includes the mid-range Terra and the higher-capability Sol tiers, giving organizations a structured way to match model choice to workload demands.

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)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Nearly Half of Top News Sites Now Block OpenAI's Crawler, Reuters Institute Finds

Major news publishers are increasingly restricting AI web crawlers, with a Reuters Institute study finding that about 48% of the world's top online news sites had blocked OpenAI's GPTBot by end of 2023. The response was sharpest in the United States, where roughly 79% of leading news sites blocked OpenAI's crawler, compared to only 24% blocking Google's AI crawler globally. Publishers are concerned that their reporting may be used to train AI models or generate answers without proper compensation, attribution, or referral traffic. High-profile actions include The Guardian blocking GPTBot in September 2023 and The New York Times suing OpenAI and Microsoft in December 2023 over alleged misuse of its content. While some publishers opt for outright blocking, others like Axel Springer have pursued licensing deals, reflecting an evolving and unsettled commercial landscape for AI access to journalism.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Apple Sign-In Breaks Every 6 Months Due to Expiring JWT Client Secret

Developers using Sign in with Apple on the web may find authentication suddenly failing with an 'invalid_client' error approximately six months after setup, caused by Apple's mandatory expiry limit on JWT client secrets. Unlike other OAuth providers such as Google or GitHub, Apple requires developers to manually generate a time-limited JWT signed with a private .p8 key, which expires after a maximum of roughly 15.7 million seconds. Apple sends no warning when the secret expires, meaning users typically encounter the failure before developers do. The standard fix involves generating a fresh JWT and updating it in the authentication provider's dashboard, but this must be repeated every six months. To automate the process, a GitHub Action called apple-client-secret-rotator has been released, which regenerates and updates the secret on a scheduled basis, eliminating the need for manual intervention.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Build Only What Differentiates Your MVP, Rent Everything Else

When scoping a minimum viable product, founders face a critical early decision: what to build themselves and what to purchase as a service. The core principle is that only the unique, differentiating feature of a product warrants custom development, while common infrastructure like authentication, payments, and email should be rented from specialist providers. Building commodity capabilities in-house wastes time that could be spent on the actual value proposition, and specialist vendors bring security, compliance, and reliability that small teams cannot replicate. However, renting carries its own risks — vendor lock-in, price changes, and API instability — so dependencies should be isolated behind internal interfaces to limit exposure. The most dangerous trap is mistaking a capability that sits adjacent to the core differentiator for part of it, leading teams to over-build where off-the-shelf solutions already exist.

0 comments Read more at DEV Community