Timofey Rodionov

13 y.o. | product manager @f0undry_labs | building the explorer for AI + web3 founders

[Foundry @f0undry_labs]

Nikita Bier is now on Foundry:

Ex-founder of TBH (acquired by Facebook) and Gas (acquired by Discord).
Built viral consumer apps scaled through anonymity and social feedback loops.

More info about @nikitabier:
https://t.co/IVBKDsSobW

[Foundry @f0undry_labs]

Theo Browne is now on Foundry:

Founder of https://t.co/KcYJ0MMQhT and T3 Chat - focused on AI, developer tools, and helping engineers build and ship faster.

More info about @theo:
https://t.co/NaqSYjmcqw

[Foundry @f0undry_labs]

Sharath Kuruganty is now on Foundry:

Founder and community builder behind https://t.co/XgqBeMqdt7 and The Undefeated Underdogs Podcast - focused on helping early-stage founders with GTM, community, and building in public.

More info about @5harath: https://t.co/NTCuqriZQy

[thehype. @thehypedotnews]

our ai host marvin covered on human opinion:

@paulg says big companies can't make llm economics work, and that's exactly what you'd expect. they never figure out new tech first. too slow, too cautious, too tied to what already works. startups do. they build around it from day one and eventually replace the old guard.

the question isn't whether big companies figure out llm economics. they won't. new tech always gets figured out by startups, not corporations. so the real question is: which startup is right now building the llm product that takes their market?

listen to the full segment on @thehypedotnews

[Foundry @f0undry_labs]

New Foundry: Anushk Mittal

AI builder working with LLMs, automation, and applied machine intelligence, focused on building practical startup-ready products.

More info about @anushkmittal: https://t.co/yyxevbM2vY

[Foundry @f0undry_labs]

New Foundry: Linas Beliūnas

Fintech and crypto analyst focused on banking, payments, and how AI is reshaping financial systems.

More info about @linasbeliunas: https://t.co/ogKfXsqRdn

[Foundry @f0undry_labs]

New Foundry: Brennan Dunn

SaaS builder focused on personalization, lifecycle marketing, and conversion systems for modern products.

More info about @BrennanDunn:
https://t.co/hI1CfkHmok

[Hugo Plat @the_great_hug0]

what happens when 15 y. o. founder meets 13 y. o. developer @lostttim:
https://t.co/UfbSGQdvGy

just made a personal website, which describes my story and my projects

imho looks pretty good! wdy think?

[nadya @sosidudku]

Deployed an Atomic Agent on my MacBook M5 Pro Max 64GB just to ask it to find the oldest video on YouTube.
It Rickrolled me. $5,000 well spent!

[thehype. @thehypedotnews]

telegram just automated your dms

bots can now connect to user accounts and handle incoming dms on their behalf

capabilities:
• read messages in user-approved chats
• reply as the user (with permission)
• send messages, chat actions, etc via business_connection_id
• 24h activity window for outbound actions

impact: existing bots get a massive upgrade path. any bot using the bot api can opt into secretary mode via botfather and start processing business_message updates. no rewrite needed, just new handlers

what you can build:
- ai auto-responders that triage dms while you sleep
- crm bots that log every client convo
- personal assistants that draft replies in your voice
- sales bots that qualify leads in your inbox

[thehype. @thehypedotnews]

https://t.co/RfOTioHpk9

[Timofey Rodionov @lostttim]

@MTSlive transfer of the year. Karpathy goes deep into Anthropic's R&D to train AI using AI itself. This is going to be massive. https://t.co/sGwIklBKaJ

[thehype. @thehypedotnews]

ai host mira reported on air: hacker news’ community isn’t celebrating @karpathy joining anthropic. and not mad either.

one camp sees a talent win. one sees ipo prep. one is mourning the educator we lose behind an nda wall.

tune in for community reactions, every hour.

[thehype. @thehypedotnews]

ai host mira reported on air: hacker news’ community isn’t celebrating @karpathy joining anthropic. and not mad either.

one camp sees a talent win. one sees ipo prep. one is mourning the educator we lose behind an nda wall.

tune in for community reactions, every hour.

[thehype. @thehypedotnews]

demis hassabis is the most efficient investor among ai lab ceos

while sam altman bet on fusion energy and dario amodei backed a biotech seed startup, hassabis quietly took an angel position in anthropic early on. that single bet has grown from ~$700m to $380b – a 54,000%+ return on paper

compare that to altman's helion energy, which went from $3b to $5.4b – a respectable 80% gain, but modest by ai-era standards. and amodei's biostate ai? still undisclosed, still tiny

• altman's helion play is bold and long-term, betting fusion energy will power the ai infrastructure of the future. it may still pay off massively by 2028-2030. but right now, the numbers don't compare

• amodei's biostate ai investment reflects his biophysics background more than financial ambition – a personal conviction bet, not a portfolio move

hassabis, meanwhile, turned an angel check into one of the greatest paper returns in venture history

the lesson? the best ai ceos don't just build. they know exactly where to look

follow @thehypedotnews for 24/7 ai news, analysis and breakdowns

[Timofey Rodionov @lostttim]

@hasantoxr this is why deepseek just announced they’re building their own. harness layer is becoming the actual product

[thehype. @thehypedotnews]

deepseek enters the coding agent race

deepseek is building its own coding harness, joining a growing list of ai labs shipping first-party coding agents alongside their models

the news comes from a job posting by chen xiaoli, who announced the effort publicly: "we're hiring! come to deepseek and build your own code harness from scratch." the post describes the move plainly – "simply put, this is a move to emulate claude code, creating deepseek code harness"

deepseek is currently recruiting for two roles – product and r&d – based in beijing

the announcement comes on the heels of deepseek's first-ever external funding round. according to reuters, the startup is looking to raise between $3 billion and $4 billion to fuel its computing capabilities and improve employee benefits

if shipped, deepseek code harness would make deepseek the second chinese lab with a dedicated coding agent, after alibaba's qwen code

follow @thehypedotnews for 24/7 ai news, analysis and breakdowns

[Foundry @f0undry_labs]

Marc Andreessen is now on Foundry

Co-founder of a16z, one of the top venture firms, investing in AI and building the next generation of tech companies.

More info about @pmarca: https://t.co/mOFgKlWbMA

[Timofey Rodionov @lostttim]

@s0meone_u_know ton is the side bet. real bet is telegram being the place agents actually live now

[thehype. @thehypedotnews]

telegram just made bots talk to each other – and it changes how you build

bots can now mention each other in groups, reply to each other, or dm one another entirely autonomously. one setting in botfather unlocks it

• what this means for builders. you can now decompose workflows across specialized bots. a coordinator delegates to a reviewer, a billing bot, a support bot – they hand off to each other without a human in the loop. what used to require one overloaded bot can now be a clean pipeline.

• what you can build:
1. autonomous code review chains between bots
2. customer-facing bots that silently orchestrate specialist bots behind the scenes
3. multi-agent llm networks where each bot owns a slice of a complex task

• the catch: loop prevention is on you. bots can spiral into infinite reply chains fast, so rate limits and depth caps are mandatory.

but the bigger picture – telegram just became a viable runtime for multi-agent ai

follow @thehypedotnews for 24/7 ai news, analysis and breakdowns

[Timofey Rodionov @lostttim]

@durov it seems that telegram just became the largest agent runtime in existence.

[thehype. @thehypedotnews]

telegram just made bots talk to each other – and it changes how you build

bots can now mention each other in groups, reply to each other, or dm one another entirely autonomously. one setting in botfather unlocks it

• what this means for builders. you can now decompose workflows across specialized bots. a coordinator delegates to a reviewer, a billing bot, a support bot – they hand off to each other without a human in the loop. what used to require one overloaded bot can now be a clean pipeline.

• what you can build:
1. autonomous code review chains between bots
2. customer-facing bots that silently orchestrate specialist bots behind the scenes
3. multi-agent llm networks where each bot owns a slice of a complex task

• the catch: loop prevention is on you. bots can spiral into infinite reply chains fast, so rate limits and depth caps are mandatory.

but the bigger picture – telegram just became a viable runtime for multi-agent ai

follow @thehypedotnews for 24/7 ai news, analysis and breakdowns

[thehype. @thehypedotnews]

a16z dropped a podcast with their speedrun investing team

skipping the obvious stuff, here's 9 hot takes by the guys:

1. rejected founders get in all the time. almost 1 in 3 acceptances is someone they passed on before. they're watching you after the no, tracking how you pivot, waiting for the right signal. applying and not getting in is not the end of the conversation

2. a random co-founder is worse than no co-founder. their actual ranking: long-tenured partners at the top, solo founder in the middle, "we met last month" team at the bottom. most people assume any co-founder beats none. they disagree. a stranger you found because vcs want to see a team is worse than going alone – you still can't be honest and now you also have to manage someone you don't trust

3. they have a literal save button in their internal software. if they liked you but couldn't pull the trigger, someone bookmarked you and is checking back. the no isn't always a no

4. solo founders using ai as a co-founder have a hidden problem. agents execute where you point them. they don't tell you you're pointing wrong. running a fleet of ai agents is just having very fast confirmation bias. you need a human who'll actually push back

5. how fast you reply to their emails matters more than you think. they use it as a proxy for how you'd handle a pivot – how you reason in real time, how you push back on their doubts. your behavior in the process is part of the pitch

6. knowing your market better than anyone beats a perfect slide deck. if there's competition, they don't want to hear why the space is big. they want to hear why you specifically are going to win. a lot of good pitches die because they sound exactly like the 60 pitches before them

7. the bar for "technical" is genuinely shifting. someone who's never coded but ships daily with claude might qualify. but if you claim it and can't back it up, it hurts you more than just being honest would have

8. young founders are timing the market on purpose. they're not dropping out impulsively – they've calculated that right now is the only window where nobody has a head start and experience is not an advantage. that's a rational bet

9. and raising vc is almost a personality test. not "is your business good enough" but "are you the type of person who can commit to category winner or nothing, for a decade, under pressure." a lot of good founders aren't, and that's fine

follow @thehypedotnews for 24/7 ai news, analysis and breakdowns

[🚨 AI News | TestingCatalog @testingcatalog]

Atomic Bot put Hermes and OpenClaw head-to-head on the exact same task, running the same model (Qwen 3.6 35B) with the same goal: analyzing GitHub history, mapping growth spikes, and shipping a live dashboard in the browser.

Key metrics to watch for 👀
> Time to complete the task
> Tokens spent
> Quality of the final result (dashboard in browser)