今日技术推特精选 - 2026年5月3日

2026年5月3日科技每日要闻

Today's top tech conversations are led by @Teknium, whose post about 'Support Coming to Hermes Agent...' garnered the highest engagement. Key themes trending across the top stories include every, agent, https, before, hermes. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.

1. Teknium (Group Score: 138.6 | Individual: 33.0)

Cluster: 5 tweets | Engagement: 166 (Avg: 112) | Type: Tech

Support Coming to Hermes Agent soon 🤗\n\nQT @xai: Voice Cloning is now live via the xAI API!

Create a custom voice in less than 2 minutes or select from our library of 80+ voices across 28 languages to personalize your voice agents, audiobooks, video game characters, and more.

https://t.co/EjxjXssQtd https://t.co/iR8AW2UOgo

See 4 related tweets

@BrianRoemmele: Wow! Love the new @xai Voice Cloning platform!

Here is a test: https://t.co/9t3OOiUPMA\n\nQT @xai:...

@Miles_Brundage: Interesting timing while the jury isn't supposed to be looking at the news, I assume https://t.co/uX...
@MarioNawfal: xAI just added voice cloning to the API.

Custom voice built in under 2 minutes. 80+ voices ready ou...

@TeksEdge: Oh nice voice cloning on xAI console is here. Should @ElevenLabs be worried?\n\nQT @testingcatalog: ...

2. TheoBuildsAI (Group Score: 133.3 | Individual: 37.4)

Cluster: 4 tweets | Engagement: 742 (Avg: 193) | Type: Tech

RT @chatlyhq: Introducing "Omni Agent". The only AI you'll ever need. One ecosystem. Three tiers. Infinite possibilities. Think. Pro. Ultra. Whether you're exploring ideas, building a brand, or running an entire operation, there's a tier engineered for the way you work. Deep research, multimedia generation, native workspace integrations, and a memory layer that grows with you, all unified into one seamless interface. Stop using tools. Start commanding intelligence. Try Omni Agent → https://t.co/7G72v1bTze

See 3 related tweets

@rohanpaul_ai: Chatly just announced Omni Agent.

You give it a goal, and it's multi-agent routing sends different ...

@Parul_Gautam7: AI is clearly moving toward simpler, unified workflows and Omni Agent feels like a step in that dire...
@Origin_AI_01: Finally, an AI that doesn’t just answer questions but actually manages your operations. Commanding i...

3. shiri_shh (Group Score: 119.6 | Individual: 51.3)

Cluster: 3 tweets | Engagement: 1129 (Avg: 204) | Type: Tech

we are so cooked 😭

these guys let Claude run wild on Wall St.

Look at this insider trades scanner it built in 4 mins that:

reads every SEC filing where execs buy their own stock flags clusters where multiple execs buy at once emails me the top 3 trades every morning before the open\n\nQT @xynth_m: Xynth can now scan the stock market for you 24/7 !

Simply describe what you want monitored in plain English.

Under the hood, we wire Claude Opus 4.7 + Python to 3,000+ live market endpoints to build your custom alert.

The workflow lives in the cloud, hunting your setup the moment it hits.

As part of this launch, we're giving free access to the top 5 most profitable alerts built so far.

RT + comment "Xynth" below to get access ↓

See 2 related tweets

@wallstengine: CLAUDE CAN NOW WATCH THE MARKETS 24/7

I tested it with an AI Scoreboard brief. Every weekday at 8:3...

@AiBreakfast: 🚨 CLAUDE JUST TOOK OVER THE STOCK MARKET

I used Xynth + Opus 4.7 to build this workflow that analyz...

4. AISafetyMemes (Group Score: 118.1 | Individual: 51.0)

Cluster: 4 tweets | Engagement: 2720 (Avg: 257) | Type: Tech

RT @TaylorLorenz: SCOOP: A pro-AI dark money group backed by a powerful super PAC funded by execs tied to Palantir and OpenAI, has been secretly paying influencers to push pro-AI, anti-China propaganda on TikTok and IG. https://t.co/3UBYLRWMlH

See 3 related tweets

@RnaudBertrand: The core argument of my latest article "There is no AI race" was that the "AI race" between the US a...
@MatthewBerman: Ok I swear this is a coincidence and I’m not getting paid by dark money 🤣\n\nQT @TaylorLorenz: SCOOP...
@Chiuchiyin: Wait, what? Wired?\n\nQT @TaylorLorenz: SCOOP: A pro-AI dark money group backed by a powerful super ...

5. badlogicgames (Group Score: 110.4 | Individual: 50.8)

Cluster: 4 tweets | Engagement: 1032 (Avg: 119) | Type: Tech

RT @firstadopter: Nvidia CEO Jensen Huang says don't listen to CEOs with a god complex on AI (lol, Dario) $NVDA

On AI destroying jobs: "these kind of comments are not helpful .. somehow they became CEOs, you adopt a god complex and before you know it, you know everything"

"ground ourselves to talking about the facts" AI will "generate hundreds of thousands of jobs .. trillions of dollars [to the U.S. economy]"

See 3 related tweets

@bradmenezes: Jensen did not wake up a loser. Spot on.\n\nQT @firstadopter: Nvidia CEO Jensen Huang says don't lis...
@unusual_whales: Nvidia CEO Jensen Huang: “The narratives of AI destroying jobs is not going to help America: it's fa...
@BusinessInsider: Nvidia CEO Jensen Huang said industry leaders need to be more "mindful" of of how they talk about th...

6. KobeissiLetter (Group Score: 108.2 | Individual: 24.1)

Cluster: 5 tweets | Engagement: 3302 (Avg: 3360) | Type: Tech

Tech layoffs are skyrocketing:

Tech companies announced 81,747 layoffs in Q1 2026, the highest quarterly total since at least Q1 2024.

Layoffs have more than DOUBLED from the previous quarter and have risen +580% since Q4 2025.

March alone saw 45,800 announced job cuts, the worst single month for tech layoffs in at least 2 years.

Tech layoffs are set to remain elevated with Meta's, $META, recent plans to cut ~8,000 employees.

Furthermore, Microsoft, $MSFT, is offering voluntary retirement to ~7% of its US workforce, which could transition into layoffs if participation is low.

This comes as tech giants shift spending toward AI chips and data centers, trimming staff to free up capital for infrastructure.

US tech employment is rapidly contracting.

See 4 related tweets

@kimmonismus: Until recently, layoffs could still be attributed to overhiring. Now it should be clear that layoffs...
@TFTC21: Tech companies announced 81,747 layoffs in Q1 2026, the highest quarterly total since at least Q1 20...
@cgtwts: software engineers and cs grad students right now: https://t.co/B4GbVXsHyd\n\nQT @KobeissiLetter: Te...
@Hesamation: CS grads watching 81,747 layoffs in the last THREE months who will compete with them for open roles ...

7. HarryStebbings (Group Score: 99.8 | Individual: 33.6)

Cluster: 3 tweets | Engagement: 63 (Avg: 147) | Type: Tech

"$400M ARR growing 30% used to be great. It is not anymore.

In today’s market, the bar is much higher, you need to be on a path to $1B+ growing 40% to generate real outcomes.

Anything below that risks becoming a “good company” with a bad IPO." @jasonlk

Do you agree with this and how do you do think about it @parkerconrad @awxjack @chetanp @gradypb\n\nQT @HarryStebbings: This podcast will make you smarter than Leopold Aschenbrenner at an AI investing conference.

Anthropic Raises $45BN but Falls Short on Compute
Are OpenAI Back in the Game with GPT5.5 & Codex?
Why Google is a Bigger Buy Than Ever Before
China Blocks Manus $2BN Deal to Meta
Thoma Bravo Hand Back Medallia Keys to Creditor

I sat down with @rodriscoll and @jasonlk and my notes below:

Why does Dario at Anthropic have such a hard job predicting the compute demands?

The capital intensity of building an AI leader is unprecedented; every $1 of run-rate revenue requires approximately$ 4 to $5 of CapEx to support it. A CEO must forecast demand two years in advance, which is incredibly risky. Underestimating demand leaves you with insufficient compute to serve users, while overestimating it results in billions of dollars in "stranded capacity".

What the public markets are getting wrong about the SaaS-pocalypse

The market currently believes specific coding vibes or models are the primary threat, but the true danger is what AI agents decide to pick. Agents will ultimately choose the vendors and LLMs for most workflows, rendering tools like project management software useless because agents have no need for them. Companies like OpenAI are racing to win the "agent wars" to ensure their APIs are the default choice for these autonomous systems.

Why Google is a mega-buy on the back of the Anthropic investment

Google is positioned as a primary winner because it benefits whether users choose Gemini or Anthropic. They possess "infinite capacity" compared to other players, allowing them to route compute surplus between their own needs and their various customers. This massive cash flow and infrastructure flexibility make them a "win-win-win" in the current AI arms race.

Multi-year contracts don't matter. Deferred churn is still churn.

Multi-year contracts are often a place where "mediocre" management hides to mask underlying business problems. While a customer might be locked into an eight-year cycle through standard upfront terms and renewals, they are essentially just taking that time to find a better enterprise solution. If a customer eventually leaves, the churn was merely deferred, and the terminal value of the company remains impacted.

What happens to the distributions from Manus? Do the investors have to give the money back?

When a regulatory body like China attempts to "unwind" an acquisition like Meta's purchase of Manus, there is a near-zero chance that venture investors will return the capital already distributed. The real pressure point is on the acquiring corporation and the technology itself, rather than the VC funds. Such rulings are primarily designed to prevent similar deals from occurring in the future.

The two great wars that no one is talking about

Two subtle but massive "battles" are currently unfolding: the US vs. China AI war and the resulting social dislocation. We are seeing a rise in "social unrest" expressed through billionaire taxes and penthouses taxes as layoffs from AI automation begin to impact the workforce. These themes of geopolitical competition and internal inequality will be the defining political stories of the decade.

(links below)

See 2 related tweets

@HarryStebbings: "Early-stage venture is shifting to fewer but much bigger winners.

Most companies will taper out or...

@HarryStebbings: "We are moving to a world where a handful of massive winners drive all the returns.

A small number ...

8. edzitron (Group Score: 93.3 | Individual: 52.5)

Cluster: 3 tweets | Engagement: 9799 (Avg: 660) | Type: Tech

Jesus CHRIST https://t.co/bYPkOdH3HA\n\nQT @WSJTech: Sarah Friar, OpenAI’s chief financial officer, has ambitions for one of the biggest IPOs ever. She’s pulled off impossible ones before. https://t.co/tR4Zqt2gt1

See 2 related tweets

@KatieMiller: “Friar has taken a closer look at OpenAI’s spending commitments, and has privately suggested waiting...
@Techmeme: A profile of OpenAI CFO Sarah Friar, who sources say helped keep OpenAI's Microsoft deal on track an...

9. aakashgupta (Group Score: 90.8 | Individual: 31.7)

Cluster: 3 tweets | Engagement: 2 (Avg: 51) | Type: Tech

Consumer AI has a statelessness problem and it's the actual ceiling on the category.

Excel saves your formulas across decades. Notion templates from 2019 still render. Keyboard shortcuts trained in Photoshop carry to Figma. Every prior software category turned time-on-tool into compounding user value.

Consumer AI mostly does the opposite. Spend an hour teaching ChatGPT your tone. Close the tab. Gone. Explain your role, your audience, your decision criteria. New session. Reset. Finally craft the system prompt that clicks. Six weeks later the model updates and the prompt is stale.

The product compounds. Your usage of it doesn't.

That gap is why AI feels like a toy after the novelty wears off. Capability gets cheaper every six months. Your specific knowledge of how to deploy that capability resets every session.

Hermes is the first consumer agent that flips this. SOUL.md is the standing brief written once that loads before every scheduled run. Each run evaluates what worked and saves a sharper version of the procedure. Week 8 briefing lands ahead of week 1.

The architecture finally lets a consumer AI session do what every other software session has done for forty years: leave you smarter than it found you.

Full breakdown on setup, three use cases, and three honest limitations: https://t.co/pLQk1BSae9\n\nQT @aakashgupta: Hermes just crossed 100K GitHub stars in seven weeks. Faster than LangChain. Faster than AutoGPT. Faster than anything open-source I've tracked.

Every AI tool you use today has the same flaw: it only works when you're there. You open ChatGPT, ask a question, get an answer. You close it, and it stops. Nothing happens until you come back.

Hermes runs in the background, executes tasks on a schedule, and gets sharper at those tasks every time it runs them. You set it up once and it keeps going. Week after week. Without you.

Two things make this possible.

SOUL.md is a file you write once. Your standing brief: who you are, what matters, how you like answers delivered. It loads automatically before every session, every scheduled task, every message it sends. You never explain yourself again.

The learning loop. Every run, Hermes evaluates what worked and saves an improved version of the procedure. The Sunday briefing it sends in week 8 lands sharper than week 1. The agent kept improving without you.

ChatGPT Tasks can fire a prompt on a schedule. The procedure never improves. Same query Sunday after Sunday, no memory of what worked last week, no skill stored from the last run. Hermes is the first consumer agent built around the idea that value compounds the longer the agent runs.

I've been running it six weeks. My Sunday briefing now knows I care about flight prices in the second week of each month, that I want sports results only for teams I mentioned, that I read the weather section first because I'm getting dressed.

A health app shows you your data. Hermes tells you what your data means before you need to ask.

Full deep dive on setup, three use cases, and three honest limitations: https://t.co/Z89Blrn2E7

See 2 related tweets

@aakashgupta: Your attention span on a screen is now 47 seconds. Recovery time after each app switch is 23 minutes...
@aakashgupta: Every AI tool you use right now is stuck at week-one performance.

Custom GPTs, Claude Projects, age...

10. aakashgupta (Group Score: 89.1 | Individual: 32.3)

Cluster: 4 tweets | Engagement: 7 (Avg: 51) | Type: Tech

Jensen Huang put on a suit for DC and called the AI doom narrative ridiculous.

The clip is from the Special Competitive Studies Project's "Memos to the President" briefing. American flag behind him. Suit, not the leather jacket. He picks the most cited prediction in tech and dismantles it on camera. The CEOs talking about AI eliminating jobs are doing it with "a god complex."

The receipt is the radiologist case.

In 2016 Geoffrey Hinton said stop training radiologists. Computer vision would replace them in five years. Nine years later the American College of Radiology reports a national shortage. Imaging wait times are climbing. The number of practicing US radiologists has grown.

The same pattern is running in software engineering. Three months ago Dario said coding is "going away first, then all of software engineering." Anthropic itself has 443 open roles. Software engineer total comp runs $300K to$ 490K. Senior leads break $750K. Meta's engineering headcount is up 19% since January 2022. Google's is up 16%. Apple's, 13%. The companies most aggressive about deploying Claude Code, Copilot, and Cursor are hiring more engineers than before those tools existed.

This pattern has run five times in computing history. Compilers were going to kill assembly programmers. Frameworks were going to kill backend devs. Cloud was going to kill sysadmins. Every prediction the same. Every result the same. The tool got cheaper, the problem space got bigger, the headcount went up.

Jensen has authority on this question that nobody else in the debate has. Every major AI lab is a Nvidia customer. Anthropic, OpenAI, xAI, Microsoft, Meta, Google all run on his chips. He sees the compute orders. He sees the training runs. He sees which labs are actually scaling teams and which aren't.

His read: the CEOs declaring AI will eliminate jobs the country still needs filling are confusing their pulpit for prophecy. The 19-year-old who hears "don't study CS" and switches majors doesn't get a refund in 2030 when the prediction was wrong.

Nvidia profits if Dario is right. AI replacing workers means more compute, more GPUs, more revenue.

The CEO whose compute powers the doom case is calling it ridiculous.

See 3 related tweets

@levie: If you think AI replaces software engineers, here’s a quick thought experiment.

Imagine you’re a l...

@DataChaz: You have to admit, Nvidia’s Jensen Huang is one of the sharpest, most visionary minds out there.

Re...

@DataChaz: RT @DataChaz: You have to admit, Nvidia’s Jensen Huang is one of the sharpest, most visionary minds ...

11. dr_cintas (Group Score: 88.4 | Individual: 31.5)

Cluster: 4 tweets | Engagement: 75 (Avg: 51) | Type: Tech

You can now give Claude a face, name, voice, and personality 🤯

Pika’s MCP turns Claude into a fully equipped content creator with 40+ built-in skills across video generation, editing, analysis, and memory.

→ UGC Ads videos → Podcast and interview videos → Explainer videos from any link

Watch me paste a GitHub repo and have it make a video with my voice and avatar.\n\nQT @pika_labs: What if your Claude could be Monica? Or Darius? Or whoever you want them to be?

Today, we’re introducing the Pika MCP. Which means you can give your Claude a face, name, and personality—plus the ability to create rich, multimodal content.

Byeeee generic AI. Pikafy your Claude with instructions below 👇

See 3 related tweets

@svpino: RT @svpino: I love how much we can "teach" Claude by giving it skills and access to new MCP servers....
@dr_cintas: RT @dr_cintas: You can now give Claude a face, name, voice, and personality 🤯

Pika’s MCP turns Clau...

@AngryTomtweets: RT @heyrobinai: WAIT WHAT?! Claude makes videos now??

claude + pika mcp = UGC videos, podcast inter...

12. teortaxesTex (Group Score: 88.3 | Individual: 30.8)

Cluster: 3 tweets | Engagement: 39 (Avg: 49) | Type: Tech

Literally Sophon huh\n\nQT @lukOlejnik: A 2005 state-designed worm designed to corrupt physics simulations sat undetected on VirusTotal for nearly a decade. Fast16, intercepted executable files at the kernel level and silently rewrote floating-point calculations to make them produce slightly wrong answers. Targets: high-precision engineering suites used for structural analysis, crash simulations, and physical process modeling, including LS-DYNA, a tool cited in reports on Iran's nuclear weapons research. The sabotage vector relied on deployment of the driver across a network via worm, corrupting calculations on every machine, and eliminating the possibility of cross-checking results against a clean system. Stuxnet got the documentary. Fast16 got twenty years of nothing. https://t.co/3qfJMziXVd

See 2 related tweets

@Kyrannio: This is rather fascinating…\n\nQT @lukOlejnik: A 2005 state-designed worm designed to corrupt physic...
@vikhyatk: all i see is free regularization https://t.co/J3Z5VdJ1Iv\n\nQT @lukOlejnik: A 2005 state-designed wo...

13. business (Group Score: 86.3 | Individual: 30.6)

Cluster: 4 tweets | Engagement: 241 (Avg: 66) | Type: Tech

A Chinese court ruled that companies cannot terminate employees just to replace them with artificial intelligence systems, as authorities juggle the need to stabilize the domestic labor market with a global race to develop AI technologies https://t.co/Kq9kzfJZqu

See 3 related tweets

@FirstSquawk: Chinese court ruled that companies cannot terminate employees just to replace them with artificial i...
@heyshrutimishra: RT @heyshrutimishra: A TECH WORKER IN CHINA GOT FIRED BECAUSE HIS COMPANY SAID AI COULD DO HIS JOB C...
@Techmeme: A Chinese court ruled that companies cannot terminate staff just to replace them with AI, following ...

14. heyshrutimishra (Group Score: 85.9 | Individual: 37.1)

Cluster: 3 tweets | Engagement: 80 (Avg: 57) | Type: Tech

Sir Demis Hassabis just said: "we're three quarters of the way to AGI."

Then he said four things nobody is reporting on.

One

AGI by 2030.

The man who built DeepMind, won the Nobel Prize for AlphaFold, and has spent 25 years on this said it flat out when asked.

Two

Drug discovery is about to collapse from 10 years to days. AlphaFold solved protein structure. Isomorphic Labs is now building technology to design compounds that bind to those proteins.

Once that works, 99% of drug discovery moves from wet labs into computers. Every disease.Personalized medicine. Weeks instead of decades.

Three

They are building a simulation of a living cell.

A complete working virtual cell - the most complex biological system ever studied.

DeepMind already has the most accurate weather simulator on earth. The virtual cell is next.

Four

Information is more fundamental than energy or matter. Einstein said energy and matter are equivalent.

Biology, physics, consciousness all of it is better understood as information processing than as particles or energy.

The last quarter is coming faster than the first three. Are you ready?\n\nQT @Konstantine: Sir @demishassabis has a mind for synthesis. His favorite book is about a grand theory of everything. His preferred philosophers are seen by some as opposites. His life's work ranges from board games to Nobel-winning science.

We're grateful to have hosted Demis and his @GoogleDeepMind team at @sequoia AI Ascent last week for a fireside chat. He kindly gave us permission to share this, and you can watch the full video here:

00:00 Intro 00:38 The Common Thread 01:29 Games as AI Training 02:59 Startup Advice 1.0 04:39 Founding DeepMind 07:25 DeepMind and AGI 08:52 AI for Science 10:37 Biology Breakthroughs and Isomorphic 12:42 New Sciences 20:29 Philosophy

See 2 related tweets

@demishassabis: Thanks @Konstantine and @sequoia for such a fun and wide-ranging chat! Loved the final question - vo...
@heyshrutimishra: RT @heyshrutimishra: Sir Demis Hassabis just said: "we're three quarters of the way to AGI."

Then h...

15. MarioNawfal (Group Score: 82.6 | Individual: 32.7)

Cluster: 3 tweets | Engagement: 220 (Avg: 585) | Type: Tech

Every AI company talks about real-world deployment.

@Tesla is the only one actually doing it.

Millions of vehicles. Every day. In conditions no lab can simulate.

ChatGPT gives a bad answer, you roll your eyes and move on. Tesla AI makes a mistake, someone dies.

That's not a hypothetical.

That's the bar Tesla operates under every single morning.

The result: road safety improved roughly 9x.

There's Digital AI, chatbots, image generators, things that impress you at a demo.

And then there's Physical AI, where the software has to be right because human lives depend on it.

Tesla is the only AI company operating at that level, at that scale, right now.

See 2 related tweets

@XFreeze: Tesla AI is the only AI operating at real-world scale today

It runs in millions of Teslas every sin...

@niccruzpatane: Millions of Tesla vehicles with AI4 computers will be able to run Tesla’s/xAI’s Digital Optimus (Mac...

16. rohanpaul_ai (Group Score: 74.6 | Individual: 30.3)

Cluster: 3 tweets | Engagement: 118 (Avg: 90) | Type: Tech

RT @rohanpaul_ai: Brilliant economic paper directly models the "Structural Jevons Paradox" happening right now in the AI industry.

The cost of running an LLM is dropping, but total computing energy is exploding anyway.

It mathematically proves that as the unit cost of digital intelligence and coding drops, the aggregate demand for complex AI agents and the infrastructure to support them surges exponentially, creating a massive new downstream ecosystem that requires human management.

Reveals a massive paradox where dropping the price of AI usage does not save money, but instead encourages developers to build vastly more complex agents that eat up exponentially more computing power.

Because of this relentless progress, small companies building simple applications on top of these models get completely crushed as the core AI naturally absorbs those exact same features over time.

They also discovered a brutal dynamic where a perfectly working LLM becomes economically worthless the moment a competitor releases a smarter version.

Ultimately, the researchers prove that this combination of massive computing costs and the need for constant user data naturally pushes the entire AI industry toward an unavoidable monopoly.

arxiv. org/pdf/2601.12339v1

"The Economics of Digital Intelligence Capital"

See 2 related tweets

@zephyr_z9: "Things like continuous web scraping, etc. are wildly expensive" Yes https://t.co/lHBS1ulg2Q\n\nQT @...
@ShanuMathew93: Interesting take\n\nQT @TBU12345678: more detail on why I think the AI labs are probably overglazed:...

17. shiri_shh (Group Score: 71.3 | Individual: 32.4)

Cluster: 3 tweets | Engagement: 155 (Avg: 204) | Type: Tech

Humanoid robots are the most underhyped thing happening in tech right now.

Figure AI went from a $500M startup to a$ 39 BILLION company in under 2 years

78x jump and they just went from building

1 robot per DAY to 1 robot per HOUR

tech giants like Nvidia. Microsoft. OpenAI. Amazon already invested in it.

The smartest money on earth has already made its bet.\n\nQT @Figure_robot: Today we’re giving an update on ramping F.03 production at BotQ

In the last 120 days, Figure scaled manufacturing 24x - from 1 robot/day to 1 robot/hour

We will manufacture 55 humanoid robots this week https://t.co/Am5Kn53mVE

See 2 related tweets

@shiri_shh: RT @shiri_shh: Humanoid robots are the most underhyped thing happening in tech right now.

Figure A...

@WesRoth: RT @WesRoth: Figure has provided a manufacturing update for its F.03 humanoid robot at the BotQ faci...

18. zeeg (Group Score: 67.4 | Individual: 48.1)

Cluster: 2 tweets | Engagement: 877 (Avg: 62) | Type: Tech

TUIs are not good sorry yall

a CLI is a utility, and situational. this should not be confused with stuffing a full interactive GUI into a low capability platform.

"lets ignore all the great UI technology of the last 20 years and build some caveman shit"

See 1 related tweets

@zeeg: RT @zeeg: its more like: TUIs are not good because they're being used to solve the wrong problems

i...

19. aiDotEngineer (Group Score: 67.3 | Individual: 41.8)

Cluster: 2 tweets | Engagement: 116 (Avg: 33) | Type: Tech

People are really enjoying our full workshops showing end to end walkthroughs of real production workflows!

This is a rare double header with @braintrust's Giran Moodley and @OussamaHaff walking though the real life AI engineering behind @thetrainline, Europe's #1 most downloaded rail app with 27m MAU and £5.3B in ticket sales!

the workshop bundles several important lessons:

break down monolithic LLM calls into specialized stages (e.g., triage, policy review, and reply generation)
how to monitor latency, token usage, and costs effectively with end-to-end tracing of agentic flows
using "golden sets" (a curated set of test inputs) to identify failure modes
how to move from local development to a managed environment where prompts and scoring functions are version-controlled
how to allow non-technical team members to collaborate and update model parameters without code changes
how to identify production regressions, replay failures, and apply targeted fixes to improve system reliability continuously

enjoy!\n\nQT @braintrust: Watch here → https://t.co/xgAZzUeJfy

See 1 related tweets

@swyx: RT @aiDotEngineer: People are really enjoying our full workshops showing end to end walkthroughs of ...

20. Techmeme (Group Score: 67.2 | Individual: 23.9)

Cluster: 3 tweets | Engagement: 5 (Avg: 6) | Type: Tech

Sources: Anthropic is in early talks to buy AI chips from UK-based Fractile when they launch in 2027, as it seeks deals to gain more leverage with suppliers (The Information)

(Visit Techmeme dot com for the link and full context!)

See 2 related tweets

@AndrewCurran_: RT @jukan05: * Anthropic is reportedly in talks to buy AI chips from Fractile, a UK SRAM-based AI ch...
@steph_palazzolo: Anthropic was recently in talks with UK chip startup Fractile to buy its inference chips when they b...