Joined 16 years ago
1. We were able to execute complete, unattended rewrites of existing systems that pass the original's unit and integration tests.
2. Low effort models were more efficient than high effort models.
3. The delivery harness and its structure had a bigger impact on consistency than model capability.
4. We saw a LOC or performance improvement on 57 of 65 ports.
Anthropic hasn't changed their licensing, just enforcing what the licensing always required by closing a loophole.
Business models aside - what is interesting is whether the agent :: model relationship requires a proprietary context and language such that without that mutual interaction, will the coding accuracy and safety be somehow degraded? Or, will it be possible for agentic frameworks to plug and play with models that will generate similar outcomes.
So far, we tend to see the former is needed --- that there are improvements that can be had when the agentic framework and model language understanding are optimized to their unique properties. Not sure how long this distinction will matter, though.
This level of throughput is achieved by including memory database within the agentic process and then the clustering system automatically shards and balances memory data across nodes with end user routing built in. Combined with non-blocking ML invocations with back pressure you get the balance for performance.
We are a pure play app dev platform and that gets to the heart of why the business model is different. I'd argue that we are very motivated to make sure that customers are successful with app dev as that is our bottom line where our rivals are financially incentives by infrastructure sales, not app dev outcomes.
We did a long podcast and a couple blogs that offered transparency to the rationale on why we moved from Apache to BSL, which still downgrades to Apache after 36 months. See Emily Omier for the specifics.
It came down to survival. The company faced a bankruptcy event as customers were using the software without contributions and after exhausting alternatives needed to change the license model to create a more sustainable approach.
The consequence of this choice was that there was less adoption from OSS and ISVs who need a flexible licensing model for embedding and redistribution. It also encouraged the Pekko fork which is a branch that is 2.5 years old. And that branch helped older projects and OSS distributions to maintain their position without financial consequences.
It is not cheap to maintain Akka, and after 15 years we have turned a profit, albeit barely. We are growing, finally, and have a prosperous future and most of our spend goes into development. It did allow us to create Akka 3, which is a simpler model for devs within enterprises mixed with a consumption based model that should be significantly cheaper than the traditional libraries, and cheaper than the cost to adopt most any other framework. We can debate the merits of different business models but we couldn't have maintained the 50 CVE fixes and create a modern version of Akka if we hadn't taken this step.
We need a better strategy on how to appeal to the OSS community once more. To appeal to startups and academics, we have free commercial licenses and subscriptions, which nearly 200 accounts have signed up in the last 18 months.
The tracking methodology buckets companies by the primary product they advertise. Withcoherence is in a different category that has a broader platform definition.
There are companies like gitlab and Codegiant that also have remote dev envs as features of the broader product line.
There are increasingly limited differentiation between various vendors. The biggest improvement areas needed now are simpler configuration, faster boot times for complex projects (pre-built code, cached artifacts, IDE plug-ins configured).
Cloud9 IDE Appvia Coder CodeSandbox CodeZero.io DevSpace Desktop Tilt Env0 Floxdev Gitpod Itopia Spaces LocalStack MetalBear Azure DevTest Labs Visual Studio Codespaces Nimbus Okteto SourcePro Porter Codeready Workspaces Repl.it Stackblitz Strong.network Subpoint Solutions Tangram.dev
I get inspired by reviewing the following things: 1. Child mortality rate over time: https://ourworldindata.org/child-mortality?country=
2. DeepMind’s protein-folding breakthrough signals a promising decade for the science of proteomics. Most directly, being able to predict protein shapes will enable us to discover drugs more rapidly.
3. The cost to produce PV modules: https://ourworldindata.org/grapher/solar-pv-prices
4. Advancement of geothermal as a potential energy source. The next generation of the industry, however, is a bunch of scrappy startups manned by folks leaving the oil and gas industry who think with today’s technology they can crack 3.5¢/kWh without being confined to volcanic regions.
5. Space exploration. The Space Shuttle entered service in 1981 and launched successfully 134 times. The payload cost to low-Earth orbit (LEO) was $65,400/kg. Today’s Falcon 9 is at $2,600/kg.
6. The improvement in adult literacy rates over time. So much more to do here, but a literate population is one that is more likely to contribute to our global productivity and success. https://ourworldindata.org/grapher/literacy-rate-adults?tab=...
7. Quantum computing experiments and trails are doubling the number of qubits every couple of years right now. Quantum computing will cause a re-imagining of security and cryptography of digital assets if it becomes production grade. https://www.bbc.com/news/science-environment-59320073
I am sure there are many other examples. Even though I am an enterprise software guy working at Dell, the progress we made in the areas of technology that we get to work in have some contributing impact to all of these trends.
