Farhan H

2026 Master list Here are some things happening this year in AI, gaming, movies & tech. AI: - Evolution of AI agents into self learning and self improvement agents. First example was @karpathy 's autoresearch and to some extent @openclaw (thanks to incredible work by Peter) with the ability to update its own files (.md) - Computers (as personal computers) have a new (alternative) definition now, thanks to Perplexity & Cursor - building on openclaw's original idea - Voice AI improvements - lower latencies and making open source models ready for some major business usecases. - Agentic applications make it to the mainstream (e.g, Google Gemini app being able to order your Uber or your groceries) for the first time. - Sneak peek of Apple Intelligence powered by Gemini at wwdc 2026 (@apple) Movies: - Nolan's next movie - Spielberg's next movie - Next Toy Story Games: - GTA 6 - Forza Horizon Japan - First major racing game set in Japan on modern day hardware - Fable - Highly immersive RPG with a lot of fun elements - Wolverine - The newest marvel classic character finally has a game (before Blade) Tech: - Next Gen Console Updates - iPhone's first ever foldable expected with no crease - first major innovation in foldables in a while since redesigning of hinges - Redesigned MacBooks expected with OLED Touchscreens - 2nd Gen of Ray-Ban Displays & further iterations on AR/XR glasses - Steam Frame, Machine & Controller - their first slate of hardware releases since Steam Deck's popularity - Phasing out of consumer PC builds as we know it due to rising prices and shortages in GPU, SSD & RAM. First year where push for cloud begins. There is a lot more I have missed that's happening or expected to happen this year and will keep updating it. Please note that these are my personal opinions as to what's expected.

@acidmodzz @araghougassian @grok What is this place

@araghougassian This too 🇨🇦🌴

if i said this is canada, would you believe me?

@julianeagu @databricks Would you sponsor a visa or support remote role?

I'm building a new team at @databricks AI Research and we're hiring. We're focused on one of the hardest open problems in AI right now: how do you measure and continuously improve agents that operate on enterprise data at scale. We're looking for founding engineers to build the flywheel that turns evaluation results directly into better agents — from development and training all the way to production. If you want to work on problems that actually matter at the frontier of AI research, I'd love to talk. Link in comments 👇

@TGGonYT @GhillieYT I think it will go live early morning and rockstar will schedule a post/video. Trailer 2 went live at 9;AM

@GhillieYT Hopefully, but a trailer first thing Monday morning seems crazy to me

Here’s how I expect Monday to unfold: - Rockstar will post nothing at their usual post time - People will crash out! - Pre-orders will go live later in the day on Monday - Trailer 3 will drop on Tuesday and will end with “Pre-Order Now”

@katiekirsch @a16z @JackSoslow @EddieKoai @mradamjafer @BaijuBhatt @handotdev @PearlHealth_ @Leona_health @phylo_bio @ethos_network @SteveStoute Anything from Canada?

just dropped a new list of open roles & breakout startups hiring right now - curated by our team at @a16z! this week featuring @JackSoslow, @EddieKoai, @mradamjafer, @BaijuBhatt, @handotdev, @PearlHealth_, @Leona_health, @phylo_bio, @ethos_network, @SteveStoute, @MarianaMinerals, @Gedmark, & more read the full list here: a16zbuild.substack.com/p/open-roles-a…

@kwindla @thinkymachines Thanks for sharing, very insightful and a lot of new things I learned.

Technical notes on the @thinkymachines multi-modal, speech-to-speech, “continuous micro-turns” demos and write-up.

@jai_mansukhanii @generalmagic_ai Already did!

@FarhanSoftware @generalmagic_ai Apply!

We're hiring at @generalmagic_ai Demand is outpacing what we can ship, and we're at the point where we need to add talent to scale. Insurance is a $7 trillion industry still running on software from another era. We're solving very complex problems in a space that hasn't seen real disruption in decades. We're building toward a world where insurance works better for everyone. Customers get faster, more personal service, brokers and carriers run more profitably, and the whole thing operates at a scale the industry has never seen before. The team is already stacked. An ML researcher out of Columbia. An engineering lead who built early software for the Ferrari Purosangue. Operators who've been at the forefront of insurance for over twenty years. Now we're looking for the next group of people. We're backed by @speedrun, @radicalvcfund, and top tier angels including @aidangomez, Kevin Wang (@Braze), @iamlarryjames and more. 10k referral bonus for any full-time hire.

The true engineers with actual engineering knowledge are still not relying on AI for 100% of their work. The rest are roleplaying as engineers with AI.

A reminder I keep coming back to: You don't need to do everything perfect 100% of the time in a year to make progress. Sometimes, we obsess over the smallest of details.

@kwindla @ekzhang1 The main bottleneck I'm trying to keep in mind is on-premise deployment. Thank you, this year will be exciting :)

@FarhanSoftware @ekzhang1 There are open options. They generally don’t perform quite as well, and using them won’t save you money for a commercial project until you are pretty large scale. But open model progress is exciting! x.com/kwindla/status…

This blog is quite good. I’ve always been annoyed by webrtc at a foundational level and it makes a good argument as to why — webrtc conflates network things (p2p, tunneling, hole punching, weird port stuff) with buffering (jitter buf, frame drops) and audio signal processing (echos, etc) across a very complex stack Like it solves a lot of problems, but I’ve too often heard even non-technical people ask “do you have webrtc” because it sounds like the only real-time thing, not realizing it’s kind of a mess anyway no one ever got fired for picking webrtc, and it’s incredibly useful, but hey client-server stream protocols are pretty good too! ain’t nothing wrong with audio packets + buffering over quic or websocket imo, even if you have to do some DSP :’) moq.dev/blog/webrtc-is…

@kwindla @ekzhang1 Cartesia*

@kwindla @ekzhang1 Deepgram, Cartesian, etc. I think you were referring to 4.3.1 and 4.4, if I understood correctly?

Thanks, I looked into it but they are closed source options, please correct me if I'm wrong. I want to build a production-grade customer service voice agent with as little latency as possible for open-source (with support for other languages). I was previously exploring Maya and researching other STS models that could be fine-tuned or tweaked with a custom pipeline if necessary. Since I'm early in my research cycle, I thought I'd ask you first.

@FarhanSoftware @ekzhang1 Latency is a solved problem. Sub-1500ms voice to voice is the norm now for a Pipecat STT -> LLM -> TTS pipeline. See that same guide I just linked to.

@kwindla @ekzhang1 What about on the model side of things? Currently, no STS models are capable enough for production usecases (or none that I know of) but maybe a custom STT and TTS pipeline that has no latency in voice responses.

Definitely use WebRTC for edge-to-cloud realtime audio. It is the best choice by far. For production scale, you should run on a commercial WebRTC provider with a global footprint, edge/mesh optimized routing, observability tooling, etc. For hobby projects, either use the SimpleWebRTC open source transport (direct peer-to-agent) or experiment with running your own WebRTC SFU server, depending on your project goals. More here: #network-transport" target="_blank" rel="nofollow noopener">voiceaiandvoiceagents.com/#network-trans…

@kwindla @ekzhang1 What would be the go-to strategy in your own experience to reduce latency for voice models? Preferably open-source

MoQ is not low latency. It was designed by streaming media people. Maybe that will get fixed. I hope so. There are definitely some things we’d do differently if we designed a new standard for low latency today. But there hasn’t been much positive movement. If you want to build your own low-latency thing on top of WebTransport, you will end up implementing a lot of WebRTC! Not all of it, but probably a lot more than you think unless you’ve done this kind of thing before. Buffering, pacing, neteq, retransmission logic, stats and observability, reconnection handling, codec-specific layers, … WebRTC isn’t perfect. But it’s complicated for a lot of real, pragmatic reasons. (The p2p stuff is only a small part of that.) This is roughly the “my database is too slow, I’m going to write my own database that doesn’t have all that fsync and transactions overhead stuff” mistake. There are also a bunch of technical mistakes in that post. NACKs are fully supported and easy to enable in browser-based WebRTC apps, for example.

@tokifyi I'd love to be a part of this.

this has to be one of the biggest builder gc on x rn reply here if you want to be part of this 🇨🇦

@gregisenberg I think history will repeat itself again just like the markets.

in a world of INFINITE content, infinite choice, infinite scroll, people are starting to want things that END - finite formats - physical products - no ai, no internet - clear boundaries there’s a real shift here and it’s going to create massive companies here we go

@lexfridman @nvidia I think it's time for a new Elon podcast for his vision on TeraFab

It was an honor to hang out with Jensen Huang, CEO of @nvidia, and do a long-form podcast with him. Really fun & fascinating technical deep-dive conversation on & off the mic. One of the most brilliant & thoughtful human beings I've ever met. NVIDIA is the most valuable company in the world by market cap and is the engine powering the AI revolution. Podcast probably out tomorrow (Monday) unless I get stuck in too many interesting conversations while running around in SF ;-) PS: I haven't checked my messages in days. Sorry for slow replies 🙏 Trying to stay deeply focused at in overwhelmingly intense time & barely hanging on. Love you all ❤️

@SawyerMerritt @SpaceX Tesla's new chip vertical is next.

Tesla D3 (Dojo 3) chip inside @SpaceX's AI satellites 😎

@elonmusk Incredible presentation. Thank you for always pushing the boundaries forward and thinking steps ahead. I am excited about seeing Tesla create its own chips and supply chain.

SpaceXAI + Tesla TERAFAB Project Goal is a trillion watts of compute/year Most must necessarily go to space, as US electricity is only 0.5TW