← Back to all episodes
Apple’s March Tease, Tiny Aya, Ring of Fire

Apple’s March Tease, Tiny Aya, Ring of Fire

Feb 17, 2026 • 7:48

Apple sets a March 4 special experience, Cohere debuts Tiny Aya for offline multilingual AI, Valve flags Steam Deck shortages, UCSF rolls out ChatGPT Enterprise, and an annular eclipse arcs over Antarctica. A fast, conversational briefing optimized for text to speech.

Episode Infographic

Infographic for Apple’s March Tease, Tiny Aya, Ring of Fire

Show Notes

Welcome to AI News in 10, your top AI and tech news podcast in about 10 minutes. AI tech is amazing and is changing the world fast, for example this entire podcast is curated and generated by AI using my and my kids cloned voices...

Quick heads up on what’s moving in AI and tech today — Tuesday, February 17, 2026.

Apple just invited press to a March 4 special Apple experience, held simultaneously in New York, London, and Shanghai. Expect budget friendly hardware — maybe a refreshed iPad, and even an entry iPhone 17e — while the big Siri AI overhauls are more likely later in the year.

Cohere dropped a new family of open multilingual models called Tiny Aya — small, fast, and built to run offline on everyday laptops.

Valve says Steam Deck availability may come and go as AI build outs soak up global memory and storage.

On the enterprise side, UCSF is rolling out ChatGPT Enterprise with HIPAA aligned safeguards.

And if you’re under southern skies, there’s an annular solar eclipse — the ring of fire — sweeping across Antarctica today, with a partial view in parts of southern Africa and South America.

[BEGINNING_SPONSORS]

Let’s start with Apple’s March 4 plans. Invitations went out for a special Apple experience at 9 a.m. Eastern — unusual, because it’s not at Apple Park, and there may be no global livestream. That suggests hands on press briefings instead.

The color blocked Apple logo on the invite hints at fun finishes, and reports line up around a push on more affordable hardware — think an entry tier iPhone 17e and a budget MacBook — while the long rumored Siri 2.0 overhaul seems more likely to surface in the spring, and again in the fall. It’s Apple, so keep expectations in check... but the timing — right before Mobile World Congress wraps — is classic momentum judo.

One more thread on Siri expectations. Reporting over the past year has repeatedly flagged delays to the more personalized, app action Siri features, pushing general availability into 2026. So, while AI will color the conversation around this event, don’t be shocked if Apple saves the heavy conversational assistant news for later milestones.

On to Cohere. The company’s research arm launched Tiny Aya — open weight multilingual models designed to run on common laptops and edge devices, not just in the cloud. The base model is roughly 3.35 billion parameters, with TinyAya-Global fine tuned for instruction following, and regional variants — Earth for African languages, Fire for South Asian, and Water for Asia Pacific, West Asia, and Europe.

Cohere says training used a single cluster of 64 Nvidia H100s — a nod to efficiency — and the models support more than 70 languages, including Bengali, Hindi, Punjabi, Urdu, Gujarati, Tamil, Telugu, and Marathi. Downloads are already on Hugging Face, with support for local tools like Ollama. If you’re building multilingual, offline first experiences — translation, on device helpers, safety critical workflows — this lands right in the sweet spot.

What’s driving the offline first push? Latency, cost, data sovereignty, and reliability. If your app needs to work on a field laptop with spotty connectivity — or you’re handling sensitive text you’d rather not ship to a remote server — models like Tiny Aya open new doors without the GPU burn rate. And being open weight means teams can inspect, fine tune, and govern behavior more transparently than with closed endpoints.

Third story — the AI boom is squeezing gaming hardware again. Valve quietly posted that Steam Deck may be out of stock intermittently in some regions, because of industry wide memory and storage shortages. The phase out of the cheaper LCD model also concentrates demand on the OLED line — which doesn’t help supply. If you tried to buy one recently and hit a wall, you weren’t alone. The memory bottleneck story is familiar by now — cloud giants and model labs are absorbing vast quantities of HBM, DRAM, and SSDs to feed training and inference clusters — leaving fewer deals and thinner channel inventory for consumer gear. Valve even hinted broader device timelines — like Steam Frame VR — could be affected by pricing and availability swings. Watch memory vendors’ guidance this quarter to see if relief is coming... or if AI build outs keep tugging the blanket their way.

[MIDPOINT_SPONSORS]

Enterprise adoption check in. UCSF flips the switch today on ChatGPT Enterprise for more than 9,000 early users, with broader campus access opening February 23. IT leadership is highlighting HIPAA alignment and higher tier data protection classifications, plus guardrails that keep institutional data out of model training. They’re also keeping their Versa API for secure access to multiple LLMs, including Anthropic, and maintaining more than 100 prebuilt assistants. Health systems are under unique pressure to capture AI’s productivity gains without compromising privacy — UCSF’s rollout is a bellwether for how large public institutions can do this in a governed, auditable way.

And finally... look up — safely. Today’s annular solar eclipse draws a ring of fire across Antarctica, with a partial eclipse visible in parts of southern South America, southern Africa, Madagascar, Réunion, and Mauritius. Peak annularity hits around 12:11:54 UTC, with a magnitude of about 0.963 — meaning the Moon covers roughly 96 percent of the Sun’s diameter along the central path. For most listeners, this will not be a dramatic sky show unless you’re far south, but here’s a helpful reminder — never view any phase of an annular eclipse without proper eye protection. If you’re a data nerd, this one is Saros 121 — member 61 of 71 in the series — lasting about two minutes and twenty seconds at maximum annularity near the centerline.

Quick recap before we go... Apple is teeing up a March 4 special experience with hardware buzz, while Siri’s bigger AI inning likely lands later in 2026. Cohere’s Tiny Aya family brings multilingual, open weight, offline capable models to everyday machines. Valve warns Steam Deck supply will seesaw as AI eats memory and storage. UCSF turns on ChatGPT Enterprise with HIPAA aligned controls. And an annular eclipse puts on a ring of fire show over Antarctica — with a partial for the far south. More tomorrow... same feed.

Thanks for listening and a quick disclaimer, this podcast was generated and curated by AI using my and my kids' cloned voices, if you want to know how I do it or want to do something similar, reach out to me at emad at ai news in 10 dot com that's ai news in one zero dot com. See you all tomorrow.