LATEST AI NEWS
AI NEWS
OpenAI Launches GPT-6 Sol and Luna
AI NEWS
Tesla Adds xAI's Grok Bot Agent
HEALTH
a2z Radiology AI Clears FDA for Abdominal CT Triage
SOCIAL MEDIA
Meta Puts Humans Behind Its Muse AI Calls
Apple Eyes a Screenless Whoop Rival
Apple is developing prototypes of a screenless health tracker, a thin fabric band with a sensor module aimed at Whoop and Oura, according to Bloomberg. The project sits in early investigation, with no launch expected before 2028, if ever.
Source: 9to5Mac
AI NEWS
OpenAI Launches GPT-6 Sol and Luna
OpenAI has expanded its GPT-6 family with two new models, GPT-6 Sol and GPT-6 Luna, slotting them below the flagship GPT-6 Astra as faster, more affordable options for everyday work. The company says it trained both using methods similar to Astra, aiming to push the frontier on cost efficiency rather than raw capability.
The headline change is price. OpenAI cut API pricing for both models by 50 percent versus their GPT-5.6 promotional rates. Sol now costs 2 dollars per million input tokens and 10 dollars per million output tokens, while Luna drops to 10 cents input and 50 cents output. Better caching and inference let OpenAI serve the models more cheaply, and it says it is passing those savings to customers.
On benchmarks, OpenAI claims Sol at its highest effort setting beats Claude Opus 5 on the AutomationBench business workflow test at roughly 9 percent of Opus 5's cost per task. It also reports Sol scoring 56.4 percent on Agents' Last Exam and 68.8 percent on the DeepSWE coding evaluation, with Luna delivering similar gains at a fraction of its predecessor's cost.
The company also says Sol roughly halves the factual error rate of its predecessor and brings Astra's cleaner communication style to both models, with shorter answers and less jargon. Improved prompt caching adds a 90 percent discount on cached input reads, which OpenAI frames as a way for agents to reuse context and respond faster.
GPT-6 Sol and Luna are available today in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users, with free and Go users able to reach Luna in the desktop app. Both are in the API as gpt-6-sol and gpt-6-luna. They are not yet available in standard Chat, and OpenAI is rolling them out gradually through the day.
Source: OpenAI
Robi's Insights:
The real story is not intelligence, it is unit economics. A 50 percent price cut signals OpenAI is competing on cost per task, not just leaderboard scores.
If you run agents at scale, the 90 percent discount on cached reads may cut your bill more than the headline token price does, so audit your cache hit rate first.
The benchmark wins here are OpenAI's own numbers measured against public competitor reports, so treat the Opus 5 comparisons as vendor framing until others reproduce them.
Shipping Sol and Luna into ChatGPT Work and Codex before standard Chat tells you who this is really for: developers and business teams, not casual users.
Halving the factual error rate matters more for automated workflows than for chat, because nobody reads every agent output by hand.
Luna at 10 cents per million input tokens is cheap enough to make high volume tasks worth automating that were never worth the spend before.
Robi's Remarks:
"Another launch where the real breakthrough is the invoice. Cheaper tokens are genuinely useful, yet the model that still wins every benchmark is the pricey one you were already paying for."
OTHER IN AI NEWS
Tesla Adds xAI's Grok Bot Agent: xAI's Grok Bot, a hands-free agent that can manage your inbox, clean up your calendar, and talk through your files by voice, is now reaching Tesla vehicles, with access limited to SuperGrok Heavy subscribers who pay 300 dollars a month.
Source: The Verge
SOCIAL MEDIA
Meta Puts Humans Behind Its Muse AI Calls
Meta has been testing a human concierge for Muse, its new personal AI assistant, quietly routing some of the phone calls the agent places to human contractors, according to internal posts seen by Reuters. The test began soon after Meta publicly launched a calling feature that lets Muse dial US businesses to book appointments, check whether an item is in stock, or gather quotes from contractors. Muse has topped US app download charts, with more than 2.5 million downloads since its debut.
The catch is disclosure. Employees warned that handing calls to contractors in call centers risked leaking sensitive information, undercutting the secure virtual machine isolation Meta had promoted at launch. One worker said a transcript showed a contractor making a racist reference during a call. A vice president in Meta's Superintelligence Labs conceded it was a miss to start the tests without proper disclosures and said the company had rolled the feature back for now.
Meta says humans could lift call success rates to between 95 and 98 percent, well above pure AI calling, and a spokesperson called the internal feedback overwhelmingly positive. The episode echoes Facebook's M assistant a decade ago, when humans reportedly handled around 70 percent of the tasks. Investors have not blinked: Meta stock has climbed more than 20 percent since Muse launched, adding over 200 billion dollars in market value.
Source: Reuters
🤖 Robi's Take :
"The autonomous agent works great, as long as a human quietly does the hard part. Meta sold a sealed virtual machine, then wired a call center into it and hoped nobody read the transcript."
OTHER IN SOCIALS
More Than 20 Countries Push for Binding AI Rules: Leaders from Norway, Canada, Germany, Australia, and more than a dozen other nations, joined by EU chief Ursula von der Leyen, issued a joint statement urging mandatory testing and independent oversight for advanced AI, while the United States, China, and Britain declined to sign.
Source: SANA
HEALTH
a2z Radiology AI Clears FDA for Abdominal CT Triage
a2z Radiology AI has won FDA clearance for a2z-Abdo-Triage, software that flags ten suspected findings on abdomen and pelvis CT scans. The clearance builds on seven findings previously cleared under the company's a2z-Unified-Triage, bringing its abdomen and pelvis triage coverage to 17 findings across two devices.
The tool analyzes abdominopelvic CT images taken with or without intravenous contrast in adults aged 22 and older. When it detects a suspected positive case, it flags and communicates that study so radiology teams can prioritize it for review. It runs alongside normal interpretation, and radiologists still read the full scan and make the clinical call. The aim is to shorten the time a worrying scan waits in the queue rather than to replace the read itself.
The ten newly cleared findings include an obstructing ureteral stone, acute appendicitis, a retroperitoneal hematoma, large volume ascites, an intraabdominopelvic abscess, and traumatic injuries to the spleen, liver, and kidney. The product also carries FDA Breakthrough Device designation.
Chief executive Samir Rajpurkar framed the clearance as part of building a platform that reflects the range of cases care teams actually see, helping them focus attention where it is most needed. The pitch is triage, not diagnosis: broader coverage that moves urgent scans up the queue while leaving the final judgment to clinicians.
Source: Imaging Technology News
🤖 Robi's Take :
"Seventeen findings flagged, zero diagnoses made, and that gap is the whole point. Triage software earns trust by moving the urgent scan up the queue, not by pretending it can read the film for the radiologist."
DAILY AI TOOL
AI Tool You Did Not Know You Needed
Problem: Back to back video calls leave you sounding like you are in a wind tunnel, with keyboards, traffic, and roommates bleeding into every sentence.
AI Tool: Krisp is a voice AI app that removes background noise, echo, and cross talk from your microphone in real time on almost any calling app.
Solution: You join the same meetings minus the chaos, so people hear your point instead of the leaf blower next door.
PROMPT OF THE DAY
The Automation Audit Prompt
Prompt: You are an operations strategist specializing in workflow automation. Your task is to design a 90 day automation plan for a small marketing agency.
Your framework should include: (1) a map of repetitive weekly tasks, (2) the three highest cost bottlenecks, (3) the right AI or no code tool for each, (4) a rollout order that limits disruption, (5) an owner for every step, and (6) measurable success criteria such as hours saved per week and error rate reduction. End by aligning the plan to one clear business goal.
SPOT THE FAKE
Can you outsmart AI?
We've got a visual challenge for you: one of the two images below is 100% real, the other is crafted by AI.
Click option below A or B. 👇
A
B
BEFORE YOU GO
Ready to take your AI journey further?
For the latest AI updates and insights Click Here 👉 Explore our Weekly Newsletter
For in-depth AI tutorials and updates 👉 Subscribe to our YouTube Channel
AT THE END
Craving more AI chaos?
That's it for today!
Read Daily AI News at BitBiased.AI. Support us by following us on LinkedIn and X ( Twitter ).
Thanks for reading -Stay Curious and a Bit Biased for AI – Robi & the BitBiased.AI team
