Today’s AI News · 2026-06-24 · 12 min
Key moments - from our scoring
Substance score
20 / 100
Five dimensions, 20 points each
Anthropic's Claude is moving from standalone interfaces into Slack itself, becoming what researchers call an "ambient AI coworker" that quietly handles tasks in the background while the new desktop pet feature - a modern, functional Clippy - provides visual feedback on progress. This represents the third major redesign of LLM UI/UX after command-line interfaces and chat windows, with implications for dozens of startups that built bridges between AI and corporate chat systems. Separately, Meta launched $299 AI glasses with Musespark AI integration, featuring visual understanding, live translation, and turn-by-turn navigation across 26 fashion-forward styles, positioning cheaper Meta-branded options against Google's high-end Ray-Bans. Most significantly, Stanford's Brian High released Proto, an open framework that acts as a universal translator across 120+ incompatible AI biology models - enabling DNA, RNA, protein, and ligand analysis to work together. Proto achieved a 32% success rate on cell-line splicing design versus the old method's 7%, with Claude already used to specify lung cancer therapies. The episode also covers rapid-fire releases including Microsoft's Maia Voice 2, Nvidia's Bio Nemo agent toolkit, Google Flow's hallucination-free video generation, and Krea's open-sourced image models, while addressing concerns from Cloudflare's CEO about AI destroying small businesses and geopolitical tensions around Meta's open-weight models.
Claude can now live in Slack channels reading context, working asynchronously without freezing the chat, and proactively nudging teams like "is this code review still happening?" when projects go quiet. Teams tag Claude like they'd tag a person, and it breaks down requests, identifies approved tools, and completes multi-step tasks in the background.
Proto achieved a 32% success rate testing 65 candidates versus the old method's 7% success rate requiring around 1,000 candidate tests - a massive efficiency gain in computational biology.
Meta's new AI glasses powered by Musespark AI offer visual understanding, turn-by-turn navigation, and live translation. They come in 26 styles including a celebrity variant with Kylie Jenner's voice, using the Meta brand for mass-market pricing while Ray-Bans remain the high-end line.
Proto acts as a universal translator across 120+ incompatible AI biology models made by different teams, allowing DNA, RNA, protein, and ligand prediction systems to work together despite previously incompatible software and formats.
By giving an invisible AI agent a physical proxy on the screen, the desktop pet provides visual progress updates while the agent works in the background, reducing anxiety by showing the machine is actively working on the task.
Our reviewer’s read on each dimension, with quotes from the episode.
This is a surface-level AI news roundup that rarely goes beyond headline summaries. The one genuinely substantive moment - the Proto biology framework stats - is presented without meaningful analysis, and the rest is filler affirmations, vague superlatives, and shallow framing.
We really are. And, um, it's completely changing how we interact with these systems.
The old method had a 7% success rate and they had to test around a thousand candidates. With Proto, they hit a 32% success rate, testing just 65 candidates.
No original thinking is present - this is pure news aggregation dressed up as analysis. Every framing device (librarian analogy, 'ambient coworker,' 'AI chessboard') is generic, and the hosts credit others (Karpathy) for the only interesting conceptual framing used.
I mean, the analogy I've been using is it's like going from walking to a library to ask a librarian a question to having that librarian just sit in on your team meetings.
He called it the third major redesign of LLM UI ux.
There are no guests whatsoever - just two unnamed hosts doing a scripted news roundup. No practitioner experience, credentials, or domain authority is demonstrated at any point in the transcript.
Welcome Back to today's AI news, brought to you by 9X Productions.
Speaker B: Yeah, her workflow is incredible.
The episode does cite real numbers in a few places - the Proto success rate improvement (7% to 32%, 1000 vs 65 candidates), the $299 price point, 15 languages for Microsoft's model - but these are just passively relayed news figures with no additional context, verification, or analytical depth.
The old method had a 7% success rate and they had to test around a thousand candidates. With Proto, they hit a 32% success rate, testing just 65 candidates.
Microsoft dropped Mai Voice 2. This is speech generation in 15 languages.
The hosts make two very brief attempts at pushback (Meta's branding, startups being displaced) but immediately accept the response without follow-up pressure. The overwhelming tone is mutual affirmation, with constant 'Right,' 'Exactly,' 'Yeah,' and 'Perfectly said' exchanges that prevent any genuine interrogation.
That's a great way to frame it.
Wow, that is a heavy but really important question to leave on.
Computed from the transcript - who did the talking, and the words that came up most.
Today we’re covering the biggest AI stories of June 24th, 2026. Anthropic launched Claude Tag - bringing the same agentic capabilities from Claude Code and Cowork directly into Slack, letting entire teams tag Claude like a coworker to handle tasks asynchronously, build context across channels and codebases over time, and even use ambient mode to follow up on conversations that have gone quiet - with Andrej Karpathy calling it the third major redesign of how people interact with AI. Meta launched Meta Glasses, a new $299 line of AI smart glasses built with EssilorLuxottica and powered by Muse Spark, spanning 26 styles including a $399 Kylie Jenner edition with a custom chime and her voice for the AI - a two-tier strategy that pairs Ray-Ban for fashion credibility and Meta Glasses for price accessibility.
Transcribed and scored by The B2B Podcast Index.
Speaker A: Welcome Back to today's AI news, brought to you by 9X Productions. Imagine hiring a brilliant new colleague, but instead of taking up a desk, they just live inside your office's chat room, instantly doing all your busy work. That's exactly what's happening today as AI moves out from standalone browser tabs and right into where your team already talks. We are officially entering the era of, um, the ambient AI coworker.
Speaker B: We really are. And, um, it's completely changing how we interact with these systems.
Speaker A: Right, so here is our mission for this deep dive. Today we are skipping the noise and looking at how AI interfaces are radically changing from Slack bots and desktop pets to, well, $299 smart glasses plus a
Speaker B: massive breakthrough in biological AI.
Speaker A: Exactly. And we'll also do a rapid fire rundown of today's tool drops and wrap up with a really heartwarming community use case. So let's jump in. You mentioned the system's changing and I think the bit biggest example of that today is Anthropic's new Claude tag for Slack.
Speaker B: Yeah, this is huge.
Speaker A: Up until now, if you wanted agentic AI, you know, an AI that can plan and do multi step tasks, instead of just typing back a response, you had to use Claude code or cowork. Now Claudtech brings that directly into Slack.
Speaker B: Right, and the mechanics of this are just fascinating because teams literally just tag the odd handle for, for a task exactly like you'd tag a real person. But, um, under the hood, it takes that request, breaks it down into different stages, figures out which approved tools it needs and just goes to work.
Speaker A: And it does all of this asynchronously, right?
Speaker B: Exactly. It doesn't freeze your chat. You don't have to watch a little loading spinner. It works in the background and then replies when it's done. But the most crucial detail is this new ambient mode.
Speaker A: Oh yeah, explain ambient mode.
Speaker B: So. So Claude basically just hangs out in your relevant channels, reading the context. If a project goes quiet, it can actually nudge you like, it'll jump in and say, hey, is this code review still happening? Andres Karpathy was talking about this today, actually.
Speaker A: Really? What did he say?
Speaker B: He called it the third major redesign of LLM UI ux. You know, going from command line to chat windows and now to this ambient presence.
Speaker A: That's a great way to frame it. I mean, the analogy I've been using is it's like going from walking to a library to ask a librarian a question to having that librarian just sit in on your team meetings. Taking notes.
Speaker B: Yeah, perfectly said.
Speaker A: But I do have to push back a little here. Like, if Claude can just live in our slack and do all the work natively, doesn't this immediately kill off, you know, dozens of agentic coworkers, startups?
Speaker B: Oh, absolutely. I mean, it puts them in a really bad spot. If your entire startup was just building the bridge between the AI model and the company chat room, well, anthropic just paved a highway. Right. Right over your bridge.
Speaker A: Wow.
Speaker B: Yeah.
Speaker A: So we have this invisible coworker now, but sometimes, you know, you don't want it to be totally invisible, you actually want to see what it's doing.
Speaker B: Which brings us to the desktop pet.
Speaker A: Yes. So for Codex, they just introduced this custom desktop pet. It's essentially like a modern clippy, but actually usable.
Speaker B: Right. Because while the AI agent is doing all that invisible background work we just talked about, this little pet sits on your screen and gives you progress updates.
Speaker A: It's so clever. And setting it up is super easy for anyone using it. You just go to Settings, then appearance and click Pets.
Speaker B: Yep.
Speaker A: Then you just prompt it, you type Use Hatchpet to create a custom Codex pet and you wake it up. But, uh, pro tip for you, if you're trying this at home, keep the design simple.
Speaker B: Yeah, you really want to stick to basic animals or creatures.
Speaker A: Right. Otherwise the spite packs get really messy if it's too complex.
Speaker B: Well. And from a psychological standpoint, giving this invisible math a physical proxy on your screen, it reduces user anxiety. We like seeing that a machine is actively working on our task.
Speaker A: So we've put AI in our chat rooms, we've put it on our desktops, but Meta is trying to put it directly on our faces.
Speaker B: Literally.
Speaker A: Yeah. Meta just announced their new Meta glasses. And the crazy part is they're starting at $299. They built these with Essler Luxottica and they come out of the box powered by Musespark AI.
Speaker B: And the hardware software integration here is really impressive.
Speaker A: Right. They have 26 different styles like Adventurer Fury, but the one everyone is talking about is the 3 Kylie variant.
Speaker B: The Kylie Jenner one.
Speaker A: Yeah, yeah. It has an embedded gem, a custom chime, and they actually use Kylie Jenner's voice for the AI. It does visual understanding turn by turn, navigation, live translation, all in the glasses.
Speaker B: And you know, Meta's strategy here is just brilliant. It's a two tier approach.
Speaker A: How so?
Speaker B: Uh, well, they're keeping the Ray Ban branding for their high end stuff.
Speaker A: Right.
Speaker B: That gives them fashion credibility. But these new ones are just branded Meta to hit that mass market cheaper price point. They are widening their lead over Google by just flooding the market.
Speaker A: I see that, but wait, I have to question the branding. Dropping the Ray Ban name makes it cheaper. Sure, but do people actually want to wear something branded Meta on their faces just to save some money?
Speaker B: Severe question.
Speaker A: I mean, even with Kylie Jenner's voice, is that enough to overcome the geek factor of wearing tech goggles?
Speaker B: Well, Meta is betting that the utility of news spark AI. Just having instant translation and visual search on your face all day will outweigh the geek factor. But yeah, the celebrity integration is definitely a calculated move to make it feel more culturally acceptable.
Speaker A: I guess we'll see if it works. But okay, let's pivot. We're using AI to look at the world around us through these glasses. But the most important AI breakthrough today is actually looking inward at our, uh, very DNA.
Speaker B: Yes, this is a massive leap for computational biology.
Speaker A: Stanford's Brian High, who created the Evo models, just released an open framework called Proto. Can you break down why this is such a big deal?
Speaker B: So the core problem right now is that we have over 120 incredible AI biology models, but you couldn't combine them
Speaker A: because they're all made by different teams.
Speaker B: Exactly. Incompatible software, totally different formats. It was impossible to use them together. Progo basically acts as a shared language across everything. DNA, RNA, proteins and ligands.
Speaker A: So it's like having 120 Nobel prize winning scientists, but they're locked in separate rooms speaking different languages. And Proto is the universal translator that finally opens the doors and lets them cure diseases together.
Speaker B: That is exactly what it is. And the test results are wild. They used it to design cell line specific splicing patterns.
Speaker A: Mhm.
Speaker B: Which is notoriously difficult.
Speaker A: Right.
Speaker B: The old method had a 7% success rate and they had to test around a thousand candidates. With Proto, they hit a 32% success rate, testing just 65 candidates.
Speaker A: Wait, 32% from 65? That's insanely efficient.
Speaker B: It really is. And they actually used Claude to specify a lung cancer therapy with it.
Speaker A: So Claude is in our slack and curing lung cancer. That's uh, that's quite a resume.
Speaker B: It gets better. Nvidia also launched their Bio Nemo agent toolkit today. It gives AI agents these callable tools for protein prediction and genitive chemistry. It really just proves that biology is the absolute hottest sector in AI right now.
Speaker A: It really is.
Speaker B: Yeah.
Speaker A: Okay, so while biology AI is making these huge long term scientific strides, the broader tech industry is just dropping tools left and right today. And facing some pretty serious geopolitical friction.
Speaker B: Yeah, a, uh, lot happened today.
Speaker A: Let's do a rapid fire rundown. Microsoft dropped Mai Voice 2. This is speech generation in 15 languages.
Speaker B: Right. And it uses zero shot cross lingual transfer, meaning you can have a single brand voice, speak French, Japanese and Arabic without needing new audio data for every language.
Speaker A: Wow. Then Baidu dropped unlimited OCR, which can process over 40 pages in one pass. And Mistral dropped OCR 4, which is layout aware.
Speaker B: Yeah, the Mistral one is cool because it actually understands the structure of a document. It knows how a chart relates to a specific subheader instead of just reading text left to right.
Speaker A: Oh, that is so helpful for financial reports. Also, Google's Flow, their creative AI studio, is now generating videos with real locations,
Speaker B: which fixes the hallucination problem. It anchors the video to an invisible 3D scaffold of an actual physical place. So the background stops glitching and shape shifting.
Speaker A: That's huge for video creators. Finally, Krea open sourced Krea 2 RAW and Crea 2 Turbo for consumer hardware fine tuning.
Speaker B: Yeah, bringing massive image generation capabilities to local laptops through extreme model distillation.
Speaker A: Which kind of ties into the broader global shift. Right. We're seeing the global AI chessboard movement. OpenAI researcher Shyamal Anadka is leaving for India to start a new venture.
Speaker B: Exactly. It proves that with these tools becoming accessible, global AI breakthroughs can happen literally anywhere now.
Speaker A: Right, it's not just Silicon Valley anymore. But not everyone is thrilled about these advancements. Cloudflare CEO Matthew Prince came out today warning that AI could, quote, destroy small businesses.
Speaker B: His logic is that local businesses survive on human persuasion. You know, their brand story, human connection. But if an AI agent is buying things for you, doesn't care about a charming origin story, it just wants the lowest price and the exact specs.
Speaker A: And small businesses just can't compete with Amazon on pure specs in price.
Speaker B: Exactly.
Speaker A: That is a really grim thought. And on the regulatory side, there is some major news today. Now, to you listening, we need to explicitly note that we are just conveying the facts here without taking any political stance whatsoever. But the Trump administration is currently pressuring Meta to submit its models for government reviews over security concerns.
Speaker B: Yes, strictly reporting the facts of the governmental scrutiny. Here it's about the tension between open source and security. Meta uses open weights models, which means anyone can download them. The administration is asking for safety reviews because while open weights drive global innovation, there's a concern they could give foreign adversaries access to highly capable intelligence Right.
Speaker A: It's a complete situation and it just shows how much governmental scrutiny is ramping up around these models.
Speaker B: Definitely.
Speaker A: So we just talk about CEOs fearing AI will destroy businesses and governments worrying about security. But let's bring it back to earth. Let's look at how this technology is actually changing real lives on the ground today.
Speaker B: I love this part.
Speaker A: We have a story from Tricia, a special education teacher in Hampstead, New Hampshire.
Speaker B: Yeah, her workflow is incredible.
Speaker A: She uses AI to brainstorm accessible lessons and to build these complex data collection systems to tabulate student success for parents. And she uses it to tailor highly specific learning plans for disabled students, students
Speaker B: which, you know, the administrative burden in special education is just massive.
Speaker A: It is. And she shared this exact sentiment with us. She said that after 30 years in special education, AI makes her feel like a first year teacher again. She is completely revitalized by the daily options to broaden her reach.
Speaker B: That is such a beautiful contrast. I mean, we spend so much time talking about AI as giant corporate efficiency or, you know, a geopolitical weapon, but here it's a profound accessibility tool. It is actively empowering individuals who are doing some of the hardest, most human jobs in society.
Speaker A: Absolutely. It's giving a 30 year veteran the energy of a first year teacher. I just love that.
Speaker B: So do I. And you know, looking at everything we covered today kind of leaves you with something to ponder. Well, think about it. If Claude is managing our workflows right inside Slack and our desktop pets are watching our code and meta glasses are augmenting our actual vision, the interface of the future isn't a screen anymore.
Speaker A: No, it's not.
Speaker B: It's just our environment. So what happens to our sense of work when we can't tell where the software ends and our human co workers begin?
Speaker A: Wow, that is a heavy but really important question to leave on. Thank you for joining us on this deep dive. If you want to stay updated on all these crazy shifts in tech, make sure you subscribe to the show. And if you enjoy today's conversation, please take a second to rate the show with five stars. It really helps us out. Have a great day and we'll see you back for tomorrow's episode.
Other episodes covering the same guests and topics, from across The B2B Podcast Index.
Claude DesignStartup to Last · on Claude Slack integration67 / 100
Building Powerful AI Image and Video Workflows for MarketersAI Explored · on Google Flow60 / 100
DX Today AI Daily Brief - Tuesday, June 30, 2026DX Today · on Nvidia Bio Nemo51 / 100
S2. Ep21 - Google's Big AI Update: What It Means for Your BusinessAI Automations for Business · on Google Flow
GPT-5.6 Sol Launches, ChatGPT Work Takes On Cowork, Meta’s Muse Spark Undercuts the Frontier77 / 100GPT-5.6 Restricted Like Fable, China Stole 28M Claude Exchanges, AI Avatar Built 200K Followers45 / 100OpenAI Built Its Own Chip in 9 Months, Fable 5 Is Coming Back, AI Is Trying to Kill the Common Cold55 / 100Grok 4.5 Rivals Opus at a Fraction of the Cost, ChatGPT’s Voice Gets GPT-Live, Seedream 5.0Meta Cooks Up a Real Image Model, Beijing Eyes AI Export Controls Too, DoorDash’s AI Report Card