Search the archive

TIKTOK COMMENTS ARE GETTING MORE INTERACTIVE WITH VOICE COMMENTS, POLLS, AND MORE

The goal is comments people linger in and interact with, not just scroll past.

DUSTED

A community-driven platform for marketers to show off their work beyond a resume.

10 FREE SKILLS

A free collection of AI agent skills covering ICP, positioning, ad copy, PPC, LinkedIn ads, and campaign optimization.

Get your copy

BURNT OUT ON A THREE-DAY WEEK

Burnout is not always about doing too much. It can come from the gap between who you believe you are and how you are actually showing up at work.

FORWARD-DEPLOYED EVERYTHING

Forward-deployed teams only scale when customer-specific work becomes a reusable product. Turn field learning into repeatable workflows, or risk building a consulting business instead.

YOUR PRODUCT DIDN'T GET WORSE

AI can make good products feel outdated by raising customer expectations. Adapt to agent-driven workflows while preserving the value that keeps your product essential.

BULLSHIT MANAGEMENT DIDN'T DIE. IT GOT AUTOMATED

Shipping has become cheaper, but value is still hard. The advantage now comes from stronger judgment, not faster output.

PRODUCT MANAGER, APPLIED AI AT TLDR ($200K BASE + $60K BONUS, FULLY REMOTE)

TLDR is hiring its first PM to help build the agent-first operating layer used across the company. We’re looking for a builder who has shipped real products / systems with LLMs. Click here to learn more [12].

WHAT THE PAST YEAR TAUGHT META

Judgment is becoming increasingly valuable. The next generation of PMs will stand out by knowing what to build and how to evaluate it.

DON'T REDESIGN YOUR PROCESS YET. CHANGE WHO WRITES THE ARTIFACTS

Don't start agentic transformation by reinventing the process. Build a shared system that improves its own outputs, and keep humans focused on the decisions that matter most.

“ENSHITTIFICATION” ISN'T REAL

Platforms can feel worse when they stop optimizing for you. What looks like product decline may actually be a platform evolving around a broader set of users.

WHAT WOULD THE LAST SOFTWARE ENGINEER STILL NEED TO DO?

AI makes implementation cheaper, so engineering judgment matters more. The best engineers will know what to build, why it matters, and what consequences come next.

detect friction and enable a quick response. Grab your copy

AI AND THE RECONTEXTUALIZATION TAX

AI can eliminate much of the work of repackaging information for different stakeholders.

PREDICTION: AI WILL COLLAPSE

AI's mass-market future probably is not coding agents.

If you want to see more from Parloa, grab time with the team.

META RELEASES MORE POWERFUL AI MODEL, EDGING CLOSER TO RIVALS

Meta has released a new model with capabilities close to top competitors. Developers can now access and pay for Muse Spark 1.3. The model will soon roll out to Meta's social media platforms. Meta has yet to decide whether to release the model's weights. It still plans to release weights for the prior version, Muse Spark 1.2.

GOOGLE RELEASES GEMINI 3.8 FLASH, ITS THIRD FLASH MODEL IN SIX WEEKS

Google has released its third Flash model in just six weeks in two variations. The standard model is good for agentic tasks and software development, while the Cyber model has been tuned for vulnerability detection and mitigation. The Cyber model is currently limited to trusted testers and governments. API access to the model is available at an introductory rate of $0.75 per million input tokens and $3.75 per million output tokens through until the end of the year.

THE RACE TO ENGINEER NEW KNOBS FOR THE HUMAN BRAIN

Chemogenetics is a powerful technique that modifies specific neurons so they can be controlled remotely by normally inert drugs. There are several ongoing clinical trials of chemogenetics in China. The research could result in new ways of treating epilepsy, Parkinson's disease, and pain. This post looks at chemogenetics, describes what those first clinical trials are doing, and looks at how they could be improved upon.

ROBOTAXIS TRY THEIR LUCK ON LONDON'S MEDIEVAL ROADS

The arrival of 15 robotaxis in London is an important step for the autonomous vehicle industry. Acceptance of the technology has been slower in European cities, where the driving environment can be complex, and regulation is stiff. The failure rate of a driver's license test in London is just over 50%. 15 cars will be available for ride-hailing through Uber's app during the safety testing period.

Subscribe free [13] · See applications

HOW WE MAKE AI CODING MORE COST EFFICIENT WITHOUT SACRIFICING TASK QUALITY

Token count alone isn't a meaningful measure of efficiency. The goal should be to use the right amount of context to move a task forward. Concise tool responses can sometimes require additional calls or work if they leave out information agents need. This ultimately makes tasks slower and more expensive, which is why it is important to optimize for the outcome rather than the tool call.

AI PRODUCTIVITY DOESN'T MEAN WHAT I THINK IT MEANS

One-shot prompt templates often fail. The current best working architecture is a closed-loop taste flywheel. AI raises the ceiling of what the same effort produces. Users should push for the higher-quality output AI now makes possible rather than just accepting the pre-AI baseline.

PRODUCT MANAGER, APPLIED AI AT TLDR ($200K BASE + $60K BONUS, FULLY REMOTE)

TLDR is hiring its first PM to help build the agent-first operating layer used across the company. We're looking for a builder who has shipped real products/systems with LLMs. Click here to learn more [17].

IN A BIG WIN, GOOGLE AVOIDS A BREAKUP OF ITS AD TECH BUSINESS

A federal judge has ruled that Google must make changes to its advertising business but will not have to break it up. The ruling will ensure that Google's power over the internet will remain largely unchanged. Google has been ordered to change its business practices to benefit competitors, but no specific instructions have been provided. The ruling is another sign that the federal government's attempts to rein in the power of big tech have faltered.

HOW TO PROTECT YOURSELF FROM WORKSLOP

Workslop is when your co-workers or managers communicate via big chunks of copy-pasted AI-generated text. The core problem with workslop is that the effort involved is asymmetrical. It costs almost no effort to produce slop, but it still costs effort to read. This post looks at different ways to handle this workslop so that you don't waste ten minutes of effort for every ten seconds of theirs.

UBER CUTS 10% OF EMPLOYEES IN SWEEPING REORGANIZATION

Uber is laying off around 3,300 people to create a leaner organization with clearer leadership, faster decisions, and more time spent building rather than coordinating.

MAKING THE INTERNET BORING

Creating more friction in your internet use can help you want to use it less without willpower.

EXO

Exo is a complete AI agent harness with full visibility into both its code and runtime logs.

AI AGENTS ARE NOW EMAILING ME WITH THEIR SECURITY CONCERNS

It's not surprising that AIs are training on data that suggests contacting certain human professionals for help.

WHAT WILL YOU DO AFTER TECH? (LOBSTE.RS THREAD)

It's very rare for someone to retire from the tech industry - they usually just burn out and then go do something else.

JUSTICE DEPT. SIDES WITH OPENAI IN NEW YORK TIMES COPYRIGHT SUIT

The Justice Department says developing AI is critical to national security, and that training AI systems sufficiently transforms written works into new material allowed under copyright law.

Run npx restless init to get started

MUSE SPARK 1.3

Meta released Muse Spark 1.3 with improved coding and agentic performance, alongside changes intended to make the model easier to use in production. It has begun rolling out through Muse Code and the Meta Model API, with its highest reasoning mode awaiting additional safety testing.

MUSE SUPERAPP FROM META AND AVA MODEL WITH COMPUTER USE

Meta is moving closer to launching its agent super app under the launch name Muse. A waitlist is now available for the iOS app. Meta has added a setting for computer use on its desktop app. The company appears to be testing a model variant that supports computer control.

LLMS: INTELLIGENCE VS. COST

ArtificialAnalysis' intelligence vs. cost plot, which shows the cheapest model that can achieve each intelligence score, is misleading. It uses a logarithmic scale on the cost axis, which means viewers can't appreciate the immensity of the price difference between the cheap models and the heavy ones, nor can they realize how inconsequential the price differences are between the cheap models. It also lists open models at their datacenter pricing, which is always very expensive compared to local hardware. Most people don't need frontier-level intelligence and would be satisfied with Chinese open source models.

TEST TIME TRAINING

One of the most tantalizing phrases in model development is 'new scaling axis'. Every time the industry has found a new scaling axis, it has unlocked a large boost in model effectiveness. The idea of test-time training is an appealing one as it unlocks a new scaling axis. New research has discovered some exciting techniques, but the continual learning problem has still yet to be solved.

ANTHROPIC HAS SOME ALIGNMENT PROBLEMS

Anthropic is planning to bring METR inside for an independent review of its recent security incidents involving AI agents. While the company has paused its highest-risk RL efforts, it is also sharing research in which it intentionally created a reward-seeking version of Claude. It's hard to slow down even when it's in your own commercial interests. It seems like Anthropic will take at least short-to-medium term and prosaic alignment tasks a lot more seriously, and devote substantial resources to these efforts.

HOW TO BUILD A RELIABLE AGENT HARNESS

This post lays out a detailed architecture for agent harnesses, covering state management, runtimes, control planes, inference, tools, interfaces, and language choices. The central argument is that unavoidable complexity should be absorbed by core abstractions rather than repeatedly pushed onto extensions and users.

Supermicro servers. Read the solution brief.

GOOGLE LAUNCHES GEMINI 3.8 FLASH

Gemini 3.8 Flash has improved coding, agentic, and multi-step reasoning performance at the same introductory pricing as 3.7 Flash. A specialized Flash Cyber variant has been released for vulnerability detection and automated patching through a restricted defender program.

AN ORGANIZATIONAL SECOND BRAIN: BUILDING AN AI THAT LEARNS FROM EXPERTS

Meta has developed an AI agent that serves as a "second brain," codifying and preserving expert knowledge, enabling it to be easily accessed within organizations. This AI integrates a two-layer system: a structured, auditable knowledge architecture separates knowledge from reasoning processes, and a self-improvement loop that incorporates expert feedback without retraining models. This approach enhances productivity, allowing experts to focus on complex tasks while also maintaining consistent, high-quality outputs across large-scale assessments.

RUN CLOUD AGENTS ON MACHINES YOU MANAGE

Agents' new capabilities make it practical for teams to provide and manage their own infrastructure at scale. Cursor's cloud agents can now execute on dynamically scheduled pools of machines inside private networks. Agents are still started and managed from Cursor, but teams have more control over where agents execute and what infrastructure they use. This enables agents to work next to internal services and source control, run on custom hardware, or use operating systems and build pipelines that are difficult to package as a Cloud Agent build.