Bayram Annakov (@Bayka) header
Bayram Annakov (@Bayka) profile picture

Product guy, systems thinker & educator. Love building stuff. Founder @appintheair. Now building AI agents for b2b sales at https://teapawt.org/t.co/St0me2s4Pv

Seattle & Bay Area
Joined December 2008
Photo by Bayram Annakov (@Bayka)Photo by Bayram Annakov (@Bayka)Photo by Bayram Annakov (@Bayka)Photo by Bayram Annakov (@Bayka)Photo by Bayram Annakov (@Bayka)Photo by Bayram Annakov (@Bayka)Photo by Bayram Annakov (@Bayka)Photo by Bayram Annakov (@Bayka)Photo by Bayram Annakov (@Bayka)
Bayram Annakov retweeted
Introducing Unreal Agent: An open-source harness with state-of-the-art cost efficiency 39% cheaper than Codex+Astra on Terminal-Bench 4.0 while maintaining performance
214
311
127
4,399
764,581
Bayram Annakov retweeted
Discovers workflow patterns and captures corrections to build permanent memory for Claude Code. github.com/BayramAnnakov/cla…
3
15
151
14,015
Bayram Annakov retweeted
Introducing the Monastery for AI-native founders. A single builder can now outperform a publicly traded company. $2 million. 12 weeks. Do the impossible.
122
144
99
1,023
367,705
Bayram Annakov retweeted
Our Applied AI team with a great write up on improving Claude’s frontend design via Skills. Also with a Claude Code plugin that packages up the skill! claude.com/blog/improving-fr…
32
118
23
1,338
461,411
Bayram Annakov retweeted
introducing openai grove: a program for early stage founders. grove builds on our existing work with openai for startups & pioneers, and includes 5 weeks of hands-on workshops, office hours, events with our team, and early access. apply here before 9/24: openai.com/index/openai-grov…
45
51
14
898
175,575
Bayram Annakov retweeted
Trailer #veo3
1
2
178
Bayram Annakov retweeted
Last week we released @OpenAIDevs gpt-realtime our latest speech-to-speech model. Prompting S2S models can be extremely powerful. To get the most out of the model we compiled a series of prompting tips of what we saw work with early customers. To present them here's Cedar our new voice in gpt-realtime walk you through it. If you want to read it yourself check the links below 👇
5
12
4
113
68,893
Bayram Annakov retweeted
🫸 STOP COMPACTING, use /clear the biggest difference I see between how the Claude Code team uses CC and the broader community how often they run into compacts instead of keeping long running chats open across multiple tasks, when you finish a task use /clear & start again
57
28
16
462
60,775
Bayram Annakov retweeted
We’ve developed Claude for Chrome, where Claude works directly in your browser and takes actions on your behalf. We’re releasing it at first as a research preview to 1,000 users, so we can gather real-world insights on how it’s used.
427
899
676
8,457
1,616,622
RT @OpenAIDevs: We've made a bunch of improvements to web search in the Responses API: 🌐 Domain filtering to focus on specific sources 📑 S…
17
10
Brian Armstrong on what he learned about management from Balaji Srinivasan “Balaji is a brilliant guy. He’s probably one of the top couple smartest people I’ve ever met in my life,” Brian begins. “He was briefly the Chief Technology Officer of Coinbase. He came in through an acquisition and did some amazing work. And he taught me how to manage a totally different type of person.” Brian continues: “Balaji is kind of unmanageable. He’s what some people might call a ‘free radical’ within an organization. He kind of bounces around, absorbing vast amounts of information — even things that aren’t his responsibility — and occasionally he would come back to me with these incredible insights.” Brian gives one funny example: “At one point he came back to me and said, ‘These are all the salespeople that are making more revenue than their salary, and these are all the people that are not.’ And the first thought I had was, ‘You’re not supposed to have access to anybody’s salary. How did you get that?’” Balaji replied, “Don’t worry about it. I found it in some database that I wasn’t supposed to have access to.” The next question Brian asked was, “How did you connect that all up?” The previous week Brian asked the data team to connect Salesforce to Coinbase’s salary data so they could start running some reports to have more accountability. But it was supposed to be a three-week project. Balaji responded, “Oh well I couldn’t sleep this weekend, and I just knew something felt off. So I had to code it up and put it all together.” When the data team completed their analysis three weeks later, they confirmed that Balaji was 100% right. “He was continually doing things like that,” Brian explains. “And he’s incredibly high in disagreeableness, which I learned from him as well. He would go into a team and ask, ‘Why isn’t this functioning well?’ And he would suffer no fools. He would not be afraid to go in there and turn half the people on a team — whether he had the permission to fire them or not… He was a very contrarian figure. I’d say about once a week someone would come into my office and say, ‘I can’t work with Balaji. He’s causing so much collateral damage.’ And I’d say, ‘Yes, but he’s also generating an enormous amount of value and I need you to learn how to work with him.’” Brian knew Balaji wasn’t going to last forever at Coinbase because it was incredibly disruptive, but ultimately he taught Brian how to be a “turnaround CEO” when needed: “In the past I was opting a little more toward trying to be liked instead of being clear about what we’re doing, where we’re going, and what the bar is. He helped me become a better CEO and have a little more disagreeableness.” Video source: @stripe (2025)
34
190
59
2,579
474,760
Bayram Annakov retweeted
In what year will the US spend more on new buildings for AI than for human workers?
49
125
39
494
187,874
Bayram Annakov retweeted
Chatted about education and stuff with @Bayka yesterday, and we brainstormed this cool idea: # Vibe learning: Posture Checks for the Mind (Thread🧵) What do yoga classes have that university seminars lack? Real-time feedback. A yoga instructor can see at a glance who's struggling with a pose and guide the entire group back into sync. But when the "poses" are conceptual—say, understanding quantum entanglement or crafting a business strategy—this immediate visual assessment becomes impossible. #AIinEducation #EdTech #FutureOfEducation #PersonalizedLearning #SocraticTutor #VibeLearning
1
1
1
1
327
Bayram Annakov retweeted
AI Markets Crystalize For subset of AI markets, it is suddenly clear who the finalists or winners will be: * LLMs * Code * Legal * Medical Scribing * Customer Service * Search & IR New cos (or incumbents) may still show out of nowhere, but at least v1 of game has crystalized
42
34
10
608
108,460
Bayram Annakov retweeted
Come to Singapore on October 3. ns.com/conference2025
192
206
54
1,726
697,916
For most users, @NotebookLM is a tool for understanding and exploring project-based information. But we also think it could be a distribution platform, amplifying expert knowledge. Today we're offering a preview of that vision: Featured Notebooks. Here's the backstory... In a way, all of this dates back to a conversation I had with @joshwoodward in late July of 2022, one of the very first I had in Mountain View after joining Google Labs. We were talking about an idea that had come to us through the work of @kevin2kelly: the concept of "intelligence as a service" that he'd written about many years before—AI that you could tap into on demand like electricity or water. I told Josh that I'd always thought that sounded like a fascinating prospect, though I’d struggled to imagine what it would look like in practice. But what we'd seen with language models, and particularly the source-grounded language models that we were starting to experiment with, had suddenly made it clear to us both how Kelly's intelligence-as-service might actually work, how it might actually give authors and publishers a new platform to share their wisdom with the world. "We're going to be able to bottle up the knowledge of experts," I said at some point. "And then people will just have that expertise on tap." Somehow the metaphor stuck: Knowledge bottles. At some point that fall, our then-designer @gabeclapper created a mock for a future page where you could purchase and download expert collections of knowledge; we called it, irreverently, "the Bottleshop." (It looked quite a bit like the design we are launching today for Featured Notebooks, as it turns out.) I kept encountering the phrase in meetings over the next year, as it spread around Labs and other parts of Google. And every time I would say: love the concept but please don't lock in on "knowledge bottles" as the official name. It was just a passing metaphor! Naming aside, the underlying premise only grew more compelling over time. General purpose models trained on aggregations of human information were incredibly useful, but imagine how much more useful they would be if they were guided by (or were guides to) knowledge that had been peer-reviewed, edited, researched by experts—knowledge that had a particular point of view. If you're looking for parenting advice, say, you don't just want the average of all parenting advice across the internet, you want parenting advice from a specialist who you trust. There was another reason why we were interested in this approach. The experience of using NotebookLM is heavily influenced by the quality of the sources you've curated in each notebook. There's literally nothing to do in the product until you load a source, and it can take time to assemble a truly rich notebook on a particular topic. Source curation (or context engineering, as we would now call it) is not something that the average user is familiar with. So being able to showcase notebooks with high-quality sources on a range of topics was important for us just to explain how the product works. So you can think of the Featured Notebooks we are launching today as a preview on two levels. For newcomers to NotebookLM, the notebooks are a preview of how useful the product can be when you've assembled a collection of sources for whatever project you're working on. But it's also a preview of a potential future where there are thousands of expert-curated notebooks on all sorts of topics that you can add to your own collection, to have the knowledge you need on tap. Our launch lineup is: Longevity advice from legendary scientist @EricTopol, bestselling author of “Super Agers” Expert analysis and predictions for the year 2025 as shared in The World Ahead annual report by @TheEconomist An advice notebook based on bestselling author Arthur C. Brooks' "How to Build A Life" columns in @TheAtlantic A science fan’s guide to visiting Yellowstone National Park, complete with geological explanations and biodiversity insights An overview of long-term trends in human wellbeing published by the University of Oxford-affiliated project, @OurWorldInData Science-backed parenting advice based on psychology professor Jacqueline Nesi’s popular Substack newsletter, Techno Sapiens The Complete Works of William Shakespeare, for students and scholars to explore A notebook tracking the Q1 earnings reports from the top 50 public companies worldwide, for financial analysts and market watchers alike In the blog post announcing the notebooks, there's a wonderful quote from @nxthompson, CEO of the Atlantic, that really captures the spirit of what we're trying to do here: "The books of the future won’t just be static: some will talk to you, some will evolve with you, and some will exist in forms we can’t imagine now," Nick said. "We’re delighted to partner with Google in its pioneering work on this front.” It's the adaptability of the notebook format that I think makes this such a compelling platform for sharing knowledge. You can read the original texts in their entirety; you can ask questions or brainstorm ideas through a conversational interface, with citations pointing you back to the relevant passages from the original sources. Students or novices can ask for simpler explanations of complex topics, or listen to Audio Overviews or review Study Guides to help them master the material. But experts can ask more challenging questions, or quickly assemble the information they need. And all of the conversational interactions can unfold in over 80 languages, no matter what language the original sources were written in. We are making these notebooks freely available to all users, either because they involve public domain information or because we have secured world rights for the material from the original authors or publishers. This is still very much an experiment, and there are many elements that we would like to improve over time. You can't generate your own studio artifacts, like Audio Overviews or Study Guides, in featured notebooks; all the artifacts have been pre-generated. Chat can be in any language, but the artifacts themselves are English-only for now. While we have tried to focus on topics that will have global relevance, the initial lineup is more U.S.-focused; if these turn out to be useful to people, we plan to diversify in terms of both language, region, and topics. The best way to get a sense of what "bottled knowledge" actually feels like in practice is to open one of these notebooks and ask for advice: ask for a sample itinerary for a 3-day trip to Yellowstone focused on wildlife, or ask for advice on a making a mid-life career change in the "How To Build A Life" notebook from The Atlantic; or ask the experts at The Economist about how global economic trends might impact your industry. I think you'll find that the results are genuinely helpful, and maybe offer a hint of a new way of interacting with an author or publisher's work. And if you have ideas for future versions of featured notebooks, please let us know!
24
64
21
449
103,315
Bayram Annakov retweeted
Mutation Mandate (short film) bio-thriller based on AI-2027 universe story by Mitya Shulick (first preview) #storyleaf #Veo3 #AIFilmmaking #AI2027 by @DKokotajlo and @slatestarcodex
1
1
1
4
936
Inception Labs has just launched the first diffusion language model publicly released for general chat Mercury is a generalist language model with similar intelligence to OpenAI’s GPT-4.1 Nano that runs >7x faster that GPT-4.1 Nano on GPU hardware. This follows @InceptionAILabs' code-focused Mercury Coder model released earlier this year. Diffusion language models achieve faster output speeds compared to autoregressive language models on the same hardware because they can process many output tokens in parallel - allowing them to leverage use more of a GPU’s compute without being limited by memory bandwidth. Mercury is available now Inception’s first party API at $0.25/$1 USD per million input/output tokens.
10
35
11
273
46,326
Two new additions to the API: 📚 Deep research 🪝 Webhooks
84
233
121
2,860
523,810